Editing Capstone final report paper Assignment | Get Paper Help

Attached is my capstone final report paper. Please help me edit it. I will also attach a presentation ppt and script to help you understand my work. You only need to edit the paper. the project is about: With recent development in connectomics, there’s a growing need for analysis of large 3D reconstructions. In this project, the students will work with a dataset acquired in P14 mouse primary visual cortex, and are expected to assist with manual annotation, learn to run deep learning models on neural data and perform analysis on objects of biological interest, we’re particularly interested in comparing the morphological statistics of mitochondria in young vs adult mouse brain, which are believed to underlie functional differences. 1. Please write an abstract section 2. Read through and edit anything you want (especially the literature review part, since initially I planned to study the comparison between developing mouse vs. adult mouse, then in fact I compared between two brain layers. please add/delete literature review to make the part consistent. The idea is to find article to support that there might exist functional difference between cortex layer4 and cortex layer 2/3 that might affect mitochondria) 3. reduce words count if you wish or add any new section if you wish 4. Please make it a better formal report paper

 

Don't use plagiarized sources. Get Your Custom Essay on
Editing Capstone final report paper Assignment | Get Paper Help
Get an essay WRITTEN FOR YOU, Plagiarism free, and by an EXPERT!
Order Essay

 

 

 

 

 

 

 

 

MScBMI Capstone Project

 

Project Title: Morphological Analysis of Mitochondria in Developing     Mouse Cortex

Name(s):

Mentor:

Scientific Advisory Committee:

Project Site:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Personal Statement

 

Presently, advancements in technologies have resulted in the use of large 3D reconstruction in different fields. As students who focus on bioinformatics, we are interested in the practical use of machine learning in real-life health science scenario. Precisely, this research aims at developing a three-dimensional (3D) reconstruction of a 14-days old mouse and finding out the morphological statistics differences of mitochondria between different cortex layers in young mouse brain. The imaging of brain structures is a topic area that has been widely explored. However, there is insufficient research and information on the morphological statistics of mitochondria in the brain of a mouse. Exploring this topic is necessary to give detailed information on the differences in the morphology of these organelles in young mice. This topic is interesting since the morphological statistics of mitochondria is considered a major factor underlying the functional differences. The research will be of great significance to learners, scientists, and researchers. It will have both scientific and academic value. The findings of this research will enable learners to have a deeper understanding of the application of machine learning in the field of biology. Specifically, these learners will comprehensively understand the concept of 3D reconstruction and its applicability to real-life science. Besides, the research will strengthen learners’ skills and understanding of how morphological statistics of mitochondria greatly differ in different cortex layers of young mice. On top of that, learners will have an in-depth understanding of the implications of such differences. Therefore, the expected personal outcome by completing this project to broaden an understanding of the differences of the morphology of mitochondria in mouse and gain necessary knowledge and skills about 3D reconstruction.

Likewise, scientific researchers, especially those who are interested in the mouse brain will benefit from the information provided in this research. The scientists will gain additional insight into the application of a 3D reconstruction in comparing the morphological statistics of mitochondria in young mice cortex layers. There is limited research on this topic, and so this current research will provide valuable information to the existing knowledge. Apart from that, this research will guide scientific researchers who intend to explore a similar topic in the future. Precisely, this research will provide in-depth and valuable information that scientific researchers can rely on to generate reliable and valid evidence concerning the morphological statistics of mitochondria in various brain layers of young mice. Furthermore, scientific researchers will base their study on this research to explore further the role of emerging technologies such as machine learning in improving science. The information on 3D reconstruction will allow scientific researchers to investigate the effectiveness of the method in imaging cells, especially those in remote regions, compared to other tools. Overall, this research will be beneficial in understanding the impact of technological advancements and the rise of machine learning in the field of science. Given the interest of further exploring the use of 3D reconstruction in viewing and understanding brain structure, this project will provide a strong foundation of similar research in the future.

 

Introduction, Background and Significance

 

Brain networks exist only at the cellular level, that is, between particular pre- and post- synaptic neurons that are connected via functionally diverse synapses situation at certain points of cell membranes (Cazemier, Clascá & Tiesinga, 2016). The concept of connectomics refers to an emerging subfield of neuroanatomy that focuses on elucidating brain networks’ wiring with cellular resolution and quantified accuracy (Cazemier, Clascá & Tiesinga, 2016). Therefore, a connectomic analysis should be capable of identifying and measuring the dendrites, axonal path, synapses, and gap junctions of the neurons involved in a specific brain network. Mitochondria are crucial organelles that are responsible for generating high-energy compounds (Sisková et al., 2010). An electronic microscopic 3D method is one of the rising strategies that are employed in identifying certain structures.

The axons and dendrites of nerve cells in the brain of mammals have a high density that light-based imaging methods can hardly resolve the entire network of neuronal wires. Following the recent development in connectomics, advances in 3D electron microscopy allow researchers to have a clear view of each nerve cell and all chemical synapses in a particular piece of brain tissue. Consequently, this increases the possibility of mapping neuronal networks densely and not just sparsely. A connectome can be defined as a connection between regions of the brain of animals (Scheffer, 2018). 3D reconstruction plays a key role in allowing researchers to view images given that it has significantly improved in terms of quality and performance. This imaging tool is useful in the automatic generation of 3D representations of objects, especially those that challenging to model. It speeds up the model generation process for graphics applications. With 3D electron microscopy, it has become possible and easy to image sizeable volumes of neutrophil. An analysis of the anatomical connections of different areas of the mouse visual cortex revealed that there is a division of the mouse cortex into two processing streams (Sporns, 2016).

According to Sporns (2016), involving high-resolution optical imaging and tracking projections across the entire brain of the mouse, has led to another mouse connectome map that charts the directed and weighted anatomical links. Overall, imaging of cells is not an easy task; however, advances in 3D reconstruction has significantly contributed to effective imaging and viewing of various cells, including the mitochondria in the brain. Due to time-consuming manual annotation and deeper development in connectomics, large 3D reconstructions for better analysis are needed. Previous studies have looked into mitochondria in mouse lung tissues or other animals such as fruit flies. As a sample of 14-days old mouse brain cortex is available, we seek to close the research gap. Even though there have been significant advances in the speed of imaging, analyzing 3D image data remains a major challenge. Considering this, dense reconstructions of cortical tissue have been limited to femtoliter-scale volumes; hence, limiting the systematic analysis of axons, neuronal cells, and their dendrites, and the dense connectome between them. Given the 3D complexity and high packing density of many structures, and the fact that axons tend to extend over a long distance to create synapses in remote brain areas, it is challenging to create connectomic maps (Cazemier, Clascá & Tiesinga, 2016). Therefore, developing a 3D reconstruction will be useful in providing an effective connectomic analysis of the mitochondria in different brain layers of young mice.

 

Description of the Problem

 

The structure of brain networks remains largely unknown, and this is attributed to technical difficulties that are involved in imaging and reconstructing the brain in 3D. The imaging of brain structures is associated with various challenges and complexities that require the use of an effective tool. In most cases, a 3D electron microscopy has been widely used by researchers to image nerve cells and synapses in the brain. Although technological advancements have led to the improvement of the performance of a 3D imaging tool, there is a need to explore how 3D can be useful in comparing the differences between brain cells. This method has been extensively used for imaging purposes; however, its effectiveness in the imaging of mitochondria in the brain of mice is not well documented. Therefore, this researcher focuses on investigating the differences in morphological statistics of mitochondria in the brain of young mouse and assessing the effectiveness of characteristic analysis in determining the differences.

In this project, the research will analyze mitochondria’s morphological features based on a dataset acquired in the P14 mouse primary visual cortex. Although several automated methods have been developed because of the requirements of the cell-tracing, those previous work makes us aware of the potential problems such as error rates and massive calculations that need scientists to have further developments. By absorbing thoughts and ways from previous work to build models and reduce merge errors, we may try our best to reduce possible errors while building reconstruction and obtaining results in our research. Our hypothesis are, Shape distribution of Layer 4 mitochondria tend to be more bipolar compared to those of Layer 2/3 mitochondria; Shape distribution of Layer 4 mitochondria tend to be more bipolar compared to those of Layer 2/3 mitochondria..  ​

Furthermore, the researchers are expected to assist with manual annotation, learn to run deep learning models on neural data and perform analysis on objects of biological interest. Particularly, this project will aim at studying a comparison of the morphological statistics of mitochondria in different layers of young mouse brain. Precisely, the researchers will create synaptic annotations and tracing mitochondria by taking advantage of a tolls called “webKnossos”. Apart from the biological importance, their texture creates some difficulty in automatic segmentation. Therefore, we wish to generate a training set from which deep learning models can learn. The researchers can then conduct the comparison of mitochondria from different cortex layers of a young mouse brain. This project is likely to add value to institutions and professionals. The professionals will have the opportunity to gain hands-on experience in data analysis and machine learning. This research aims at enhancing professionals’ understanding of computational biology, develop skillsets for our next career stage, and find some functional differences between brain layers of young mouse. On top of that, institutions such as research centers, information technology organizations, and healthcare facilities will enormously benefit from this research. For instance, the research centers will have access to the additional information provided by this research regarding the connectomic analysis of mitochondria in the brain of young mice. Besides, the research focuses on exploring the practical use of machine learning in the field of biology, particularly covering 3D reconstruction. In this regard, information technology organizations will be well-informed on the effectiveness of the application of 3D reconstruction in health sciences. Consequently, they will have an idea of areas that need improvement and even design new strategies for better imaging of brain structures. Further, the findings of this research provide healthcare institutions with an opportunity to understand the current evidence about the morphological differences of mitochondria in brain layers of young mice. Generally, this project is more likely to be important to different stakeholders.

 

Literature Review

 

While extant studies examine the neuronal exchanges in the brain cortex, the broader organization of the communication network across the cortex remains unclear. Yet, these networks underlie the expression of cognition and behavior. This expression requires an integration of multiple sensory inputs, harmonization of several motor outputs, and synchronization of activities within the system that connect the two (Zing et al. 2015). Consequently, researchers must develop a brain-wide connectivity map for all regions of the gray matter, which will enhance understanding of the neural influences on behavior and cognitive function.

Several automated techniques have been used to probe the structure of the neuronal cortex. Firstly, Kasthuri et al. (2015) explored the structure of the neural tissue using automated techniques. The researchers opined that reconstruction effort of the mammalian brain is a costly undertaking, and substantial savings would be produced by examining overlaps of dendrites and axons to infer connectivity. However, reconstruction of an entire mammalian brain would result in a large and superfluous dataset, which would present significant challenges in processing. Consequently, Kashturi et al. used a small piece (1,500 µm3) of neocortical tissue to reconstruct all the connectivity within the cortex. Using saturated reconstruction, the researchers identified each point of contact between excitatory axons and dendritic spines. They found that nine different axons appose each spine, and only a single axon institutes a synapse. Kasthuri et al. offer sufficient explanation of the complex nature of mapping the neuronal cortex. Nonetheless, the research provides a simplified illustration of mapping the brain cortex and can be replicated on a larger scale.

Changes in neuronal connections influence experiential learning. A study by Dorkenwald et al (2019) presented a large map of connectivity between cortical neurons using electron microscopy, which demonstrated the learning processes used by the cortex. These researchers demonstrate that synapse size can be modelled using a binary and analog variable derived from a log-normal distribution. Using data from a mouse’s primary visual cortex, the research finds that synapse size equates to a binary variable combined with a log-normal continuous variable. Further, the research finds that synaptic plasticity-or the biological processes that underlie learning and memory- modifies these binary and log-normal variables. \

Similarly, Motta, Alessandro, et al. (2019) raised the efficiency of these methods by improving the automated segmentation quality, analyzing the automated segmentation for locations of likely errors and directing the human work to these locations and optimizing human data interaction by helping annotators to immediately understand the problem to be solved.

Mitochondria are important organelles for brain cells as they facilitate energy production in the form of ATP, thus allowing the brain neurons to send messages and coordinate body functions via synaptic transmission, calcium uptake, which allows electrical signals to be transmitted between nerve synapses, and the catabolism of reactive oxygen species (ROS), which are harmful to brain cells. Hence, the level of activity in a brain cortex determines the number of mitochondria found therein and their size. They are transported from one region to another via the axons and dendrites of nerve cells. The mouse cerebral cortex is made up of six layers with a uniform laminar structure. Layers 1, 2/3, and 4 constitute the upper layers and process information locally, which is attributed to their localized intracortical connections, while layers 5, 6, and 6b form the deep layers with longer connections to subcortical targets, such as the spinal cord and thalamus, among others, and the opposite hemisphere (Belgard et al., 2011). The number of mitochondria in the upper layers of the mouse brain cortex increases progressively from layer 1 to layer 4, with the highest volumes located in the dendrites of layer 2/3 to facilitate the intracortical connections between the layers. However, most neuroscientists use rats as it is difficult to analyze the mouse brain cortex accurately due to the small amount of tissue available.

In their study of the mitochondria found in the somatosensory cortex of the juvenile rat and the distribution of the organelles between nonsynaptic, axonal, and dendritic processes, Santuy and colleagues (2018) found out that the volume fraction of mitochondria in layer 4 was 7.74% in comparison to the 4.59% observed in layer 1, with statistically significant disparities between layers 4 and 2/3 and all the other layers. The mitochondrial volume was also highest in the dendrites of layer 3 at 48.70%, compared to layer 2 at 46.50% and layer 4 at 43.92%, and at the excitatory axons of layer 3 (14.35%), relative to layer 4 at 13.53% (Santuy, Turegano-Lopez, Rodriguez, Alonso-Nanclares, DeFelipe, & Merchan-Perez, 2018). These differences show that the highest energy demand levels for dendrites are at the connection between layer 2/3 and 4, primarily because layer 2/3 is at the middle of the upper cortex. However, considering all the mitochondria in the dendrites, excitatory and inhibitory axons, and nonsynaptic elements, the volume fraction of the organelles was highest in layer 4 at 7.74%, compared to layer 3 at 6.40% and layer 2 at 5.22% (Santuy et al., 2018). Therefore, it can be deduced that there is a positive correlation between mitochondrial volume fraction and synaptic density and activity. Layer 4 of the cerebral cortex has the most mitochondria because it receives long-distance thalamic afferences with high firing rates to facilitate efficient transmission of signals (Santuy et al., 2018). The mitochondria in the dendrites also had distinct morphologies to those found in the axons. Dendritic mitochondria were in high volumes and with long, tubular shapes due to the high degree of fusion in the region, whereas those in the axons were small and punctate due to their high levels of fission (Lee, Hirabayashi, Kwon, Lewis, Jr., & Polleux, 2018). This unique morphology influences the function of the mitochondria at vesicle release sites as the long, tubular shapes promote increased ATP synthesis to pump sodium ions out of cells as the post-synaptic side of layer 4 is more concentrated with the ions compared to the presynaptic side.

Brandt et al.’s (2017) study finds that age-related changes in mice mitochondria were more pronounced in heart and liver cells. The researchers explore the impact of ageing on mitochondrial function and ultrastructure in fruit flies and mice. They find that the mitochondrial ultrastructure and function remain unchanged in heart cells, while liver and kidney cells show age-related alterations in membrane morphology. Consequently, the study establishes a clear link between functional decline of mitochondria and the architecture of the inner membrane. Further, they observe that these changes depend on the organism, as well as the extent to which the tissues are protected against damage. The study offers important insights on the effects of ageing on animals with different lifespans.

Brandt et al.’s study used a more representative sample of mice and fruit flies. The sampled mice ranged from 20 to 96 weeks of age. The study estimated the average lifespan of the mice as 107 weeks for males and 114 for females. Further, the study used high-resolution respirometry and electron cryo-tomography to assess mitochondrial function, thereby enhancing the accuracy of measurement. Consequently, the researchers identified abnormal morphologies in mitochondria including widened crystae, crystae membrane swirls, voids, granules, and membrane enclosures (Brandt et al., 2017). However, the research did not examine brain cells, which play an important role in the biogenesis of the mitochondria. Nonetheless, the expanded age-range of the sample offered an expanded view of the changes in morphology over mice’s lifetime.

Developing a full-scale map of the brain-wide neuronal circuitry presents major challenges for researchers. The reconstruction effort is complex and costly and results in large datasets that add to the complexity. Yet, understanding the structure and connections of the brain helps in analyzing the cognitive and behavioral characteristics of the subjects. Recent developments in imaging technologies and algorithmic processing have improved the reconstruction modelling capacity, but major hurdles remain in integrating large data sets to develop complete images. While the technologies are still evolving, major hurdles remain in developing full-scale brain models needed to study the nature and size of synapses, and their influence on cognition and behavior. The reviewed studies reflect a gradual development in the right direction but leave significant research and practice gap. Therefore, this research seeks to close this evident research gap.

Description of the project and anticipated results

 

This section describes the methodology and timeline of the project, from inception to completion. The major steps of the project include:

  1. Reading papers and get knowledge of previous work
  2. Using WebKnossos to trace mitochondria in samples of different parts of young mouse brain
  3. Validating manual tracing accuracy
  4. Using Unet to finish tracing automatically the complete datasets from layer 4 and layer 2/3
  5. Analyzing the morphological statistics differences of mitochondria from two layers.

 

Various methods can be used in tracing mitochondria, including webKnossos which is an automated method, manual tracing, and UNet. In this study, the researcher intends to use the automated methodology, WebKnossos to significantly trace mitochondria. Tracing and analyzing as the major steps in this study that requires much attention. The major risk associated with this study that is worth addressing is that there may be some errors during manual tracing of mitochondria. Thus, in order to validate manual tracing accuracy, two ways of quality testing are desired. First, the tracing task will be divided and traced by two students individually. Thus, works can be examined by each other. Second, once manual trace is done, the piece will be divided into two potions, training set and testing test. Machine learning will be implemented to validate accuracy. Once manual tracing is done, it can be uploaded to UNet, and used as training dataset for future automated tracing.

Additionally, analysis is essential and should provide necessary information regarding the main contribution of the project. The phase of analyzing image data will be mainly completed using python. Possibly, the targeted measurements may include the 3D shape, volume size, spatial distribution, etc. With that, further analysis will be based on comparison with mitochondria in different mouse brain layers. The researchers will be using Jupyter Notebook, a python shell, and two handy python packages, numpy and ndimage. The idea is to transfer image data to digital data for numerical analysis.

The problems, especially merge errors that may occur resulting from the automated tracing of mitochondria is related to fragmentation of the organelle which is an indication of unhealthy cells and is caused by high scanning laser power, and high energy and low wavelength laser. Furthermore, damage to the tissue sample could lead to break of cell membrane, and carbon dust could result in the appearance large cluster of black matter. Another problem is immotile mitochondria resulting from high speed of imaging. These factors interfere with machine tracing; hence, affecting the overall outcome. Afterall, p14 mouse brain tissue is fragile. However, with proper manual tracing, machine can be trained to avoid the error. The equipment or software needed to conduct the study is laboratory computers, particularly, a supercomputer is necessary. When it comes to costs, there is no need for additional costs.

The anticipated results of the morphological analysis of mitochondria will be presented in diagrams and tables of analytic results. The outcome will be based on comparing the morphological statistics of the mitochondria from different brain layers. Since the brain is highly sensitive to the morphological changes caused by aging, it is possible that the synaptic mitochondria of young mouse won’t show much difference due to its young age. Nevertheless, our project results will contribute to Kasthrui’s lab, as well as to the field of connectomics.  Overall, the anticipated outcome of the analysis will be presented in tables, or diagrams appropriately.

The project’s benefits to the sponsor organization or research field is that manual tracing will assist in alleviating some of the problems associated with machine learning workflow. Likewise, the morphological analysis of mitochondria will enable individuals to gain more biological outcomes from the datasets. Therefore, this project will be beneficial to both researchers and sponsor organizations.

 

Project Methodology — Tracing

The purpose of tracing is to identify and locate mitochondria in brain cells. This step is fundamental since only accurate tracing could lead to significant results, otherwise the results would just be meaningless. This section shows the tracing phase of our project in detail.

The image above is the process of manual tracing, we identify mitochondria with eyes and paint them with different color. This process can be very tedious and time consuming since there are thousands of mitochondria in our dataset, so we would hope machine can do the tracing work for us. However, since the field of connectomics is relatively new and suffers from shortage of high-quality training data, automated tracing could be problematic at this point. Therefore, we traced manually a portion of the data to create training sets to bootstrap different machine learning models. We cut roughly 3 percent of each complete data as our sample dataset and assist those with manual annotation. The final number of mitochondria we traced in sample dataset manually are layer 4 with 288 mitochondria, and layer 2/3 with 200 mitochondria. Then we used the manual traced results as train set for their complete layer dataset separately, sample layer 4 for the complete layer 4 dataset, while sample layer 2/3 for the complete layer 2/3 dataset. After a test run, we noticed some training error shown below.

 

Training errors are identified at the bottom left as well as the top right corner along the edges of cell membrane. Machine falsefully labeled some areas as mitochondria. This can be caused by color similarity which can be misleading to the machine. We manually fixed those training error before implementing the algorithm on the complete datasets. ​ After all, our complete annotated layer datasets are:

Layer 4: five gigabytes in file size with 6948 traced mitochondria

layer 2/3: five gigabytes in file size with 6837 traced mitochondria

 

Project Methodology – Analysis

 

To study features of mitochondria of biological interests, we converted brain cell image data into the forms of matrices for numerical analyses. This section explains the methodology for each steps of the analysis. For a clear illustration purpose, the section uses the sample dataset, sample set from layer 4 with 288 mitochondria.

The first step was to study mitochondria’s size, we calculated the mean and the standard. The calculated for mean for layer 4 sample mitochondria was 44257 while the standard deviation was 49590. The mean and the standard deviation would be essential for comparing of the results to determine if the two layers are comparable. To ensure that we got the correct results from the comparison of the mean and the standard deviation, we used histograms as shown in the figures below.

 

Our interest was also in determining the length and the direction of the mitochondrion, in order to do that, we performed PCA, principal component analysis, on each mitochondrion to get three principal component and its growing direction. The plot below is the result of sample layer 4, each cone shape is a representation of mitochondria.

 

However, since the axis is normalized in terms of length during PCA process, we decided to transform back the axis to their physical coordinates. After that, we got the length of the three essentials axis for each mitochondrion. Those axis are similar to the length, width, and height of a cube. Nevertheless, since mitochondria are ellipsoid in shape, we got one long axis and two short axis. We defined the long axis as its pointing and growing axis, because in our physical interpretation, it would seem like the mitochondrion is pointing at its foci; and when viewed from a mathematical dimension, the long axis outgrows the other two axis. Below are the length distribution graph of one growing axis and two short axis.

Growing axis:

Short Axis 1:

Short Axis 2:

 

The three axis were then combined in one graph for the difference to be studied and later we combined the three axis into one distribution graph to ensure our findings are correct.

Another strategy that we considered, and we carried out was on k-means clustering based on direction. This was to help us determine the direction that each mitochondrion took during their growth. We agreed that each arrow would be representation of a mitochondria and the clustering was according to their orientation. We faced a slight challenge in determining the head and the tail of foci since the shape of the mitochondrion is ellipsoid. However, this did not hinder us from continuing with our experiment. Since we never wished to stop halfway, we considered conversion of each mitochondrion dimension coordinates to ensure that it is possible to compare. This made it possible to determine the foci according to the direction that each took. We classified the data points into five groups. Each color represents a group, and each arrow in one group points approximately the same direction. This is clearly shown in the diagram below.

 

The number of mitochondria in each cluster: [113 77 89 67 59].

 

At the end, we performed length distribution of growing axis based on k-means clustering. For us to succeed, we considered using the results we had obtained from k-means clustering. This would help us in studying the growing patterns of the mitochondrion.

 

 

The length distribution of growing axis based on k-means clustering result with distribution curve is as follows:

 

Preliminary Findings and Results

 

After implementing codes on the completed datasets, the results are shown and compared in this section. For size distribution of mitochondria from the complete layer 4 set, the mean is 18345, and standard deviation is 34965. While size distribution for the complete layer 2/3 set, mean is 16597 with standard deviation of 24730. As shown in the graph, the graph does not start exactly on zero on the x axis. This is because machine labeled a fair amount of dust segment as mitochondria, therefore, we set the minimum threshold as 2000 nanometer.

 

Size distribution

Layer 4

(Mean: 18345, std: 34965)

 

Layer 2/3

(Mean: 16597. Std: 24730)

 

Comparison result: As shown in the histograms, the mean from two size distribution graph is similar, both have very large standard deviation meaning the points are far away from the mean. Nevertheless, standard deviation from layer 4 is greater. This is because layer 4 contains more extremely small mitochondria, as well as more extremely big mitochondria. While layer 2/3 mitochondria tend to cluster at a lower middle size. However, the total number of those extreme ones in layer 4 are not significant enough to change the bell curve. From the graph, the bell curve looks similar, both extremely right skewed, meaning the majority of mitochondria from both datasets are small ones. The extremely big ones from layer 4 makes it curve move slightly a bit towards right.  So, we cannot conclude that size of mitochondria is brain layer 4 are more bipolar distributed compared to those from layer 2/3. Overall, the size distribution is similar.

 

PCA analysis

Layer 4:

 

Layer 2/3

Comparison result: It is difficult to find difference based on visual comparison since the dataset is huge. At this point, they look similar, we will quantify the characteristics for further analysis.

 

Length Distribution of Growing/point axis

Layer 4

Layer 2/3

 

Length Distribution Short Axis 1

Layer 4

Layer 2/3

 

Length Distribution of Short Axis 2

Layer 4

Layer 2/3

Combined Length Distribution of Three Axis

Layer 4 

Layer 2/3

Comparison result: Both mitochondria from brain layer 4 and layer 2/3 displaced a bell curve of gaussian distribution. However, layer 4 has shown numbers that are larger in the two short axis, and smaller in the growing and pointing axis compared to those form layer 2/3. This means that mitochondria from layer 4 tend to be more sphere-shaped, while those from layer 3 tend to be ellipsoid-shaped. Nevertheless, the difference is not much significant.

 

K-means Clustering Based on Orientation

Layer 4

 

Number of mitochondria in each cluster: [1515 1457 1347 1365 1264]

 

Layer 2/3

 

Number of mitochondria in each cluster: [1235 1435 1362 1179 1626]

Five Approximate Direction in 2D

 

Comparison result: The number of mitochondria in each cluster from the two brain layers are similar. Each cluster has more than 1200, and less than 1600 mitochondria. This means, although there are certain directions that have slightly more mitochondria pointing, there exists no dominant direction that all mitochondria tend to point at. We performed a length distribution of growing axis based on k-means clustering results in order to study patterns in growing direction.

 

Length Distribution of Growing Axis Based on K-means Clustering Result

Layer 4

Layer 2/3

 

Comparison result: Overall, mitochondria from the two brain layers have very similar patterns in orientation. From the graph, for both brain layers, mitochondria with long growing axis tend to point in the purple direction. Mitochondria with short growing axis tend to point to the green direction. The length of growing axis from layer 4 mitochondria is comparable with the length of those from layer 2/3 in both the green and purple direction. This represents that mitochondria pointing to the purple direction are more likely to outgrow those pointing in the green direction. The difference is that, in the orange and red direction, layer 4 contains larger mitochondria compared to layer 2/3. While in the blue direction, layer 2/3 consists of larger mitochondria compared to layer 4.

 

Conclusion and limitation

 

Overall, morphological statistics of mitochondria between brain cortex layer 4 and brain cortex layer 2/3 of developing mouse exist no significant differences. However, few minor variations can be identified based on our study. First of all, layer 4 contain slightly more mitochondria, and their size tend to be more bipolar compared to those of layer 2/3, namely extremely large and extremely small ones. Mitochondria from layer 4 tend to be more sphere-shaped while those from layer 2/3 tend to be more long-ellipsoid-shaped. Although there is no dominant direction that mitochondria are pointing at for each layer, there are small differences in size between each clustering direction.

The only regret left from this project is that we did not confirm any of our hypothesis, since there exists no significant morphological difference. This could because that the mouse is too young to show significant difference in the brain layer, or simply mitochondria are similar between the two brain layers. We are positive that more information can be inferred if compared to the same layer region of an adult mouse brain. Nevertheless, this project is significant in the way that with the same methodology and the same codes, we can analyze and compare morphological difference of mitochondria from any parts of any mammal, if we have the specific data set available. For instance, if one wishes to study the morphological difference of mitochondria between dog’s lung tissue and its brain tissue, he can simply go through the manual tracing process, implement the same machine algorithm, and change the data input in the coding. Moreover, after accumulation of accurate training tracing data, the stage of manual tracing will no longer be necessary. Much more information can be drawn, and further studies can be conducted based on this project.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

References

Belgard, T. G., Marques, A. C., Oliver, P. L., Abaan, H. O., Sirey, T. M., Hoerder-Suabediseen, A., … Ponting, C. P. (2011). A transcriptomic atlas of mouse neorcortical layers. Neuron, 71(4), 605-616. doi: 10.1016/j.neuron.2011.06.039

Brandt, T., Mourier, A., Tain, L.S., Partridge, L., Larsson, N., & Kühlbrandt, W. (2017). Changes of mitochondrial ultrastructure and function during ageing in mice and Drosophila. eLife Journal, 6, 1-19.

Cazemier, J. L., Clascá, F., & Tiesinga, P. H. (2016). Connectomic analysis of brain networks: Novel techniques and future directions. Frontiers in Neuroanatomy, 10, 110. https://doi.org/10.3389/fnana.2016.00110

El-Merhie, N., Baumgart-Vogt, N., Pilatz, A., Pfreimer, S., Pfeiffer, B., Pak, O., Kosanovic, D., Seimetz, M., Schermuly, R.T., Weissmann, N., Karnati, S. (2017). Differential alterations of the mitochondrial morphology and respiratory chain complexes during postnatal development of the mouse lung. Oxidative Medicine and Cellular Longevity, 1, 1-22.

Januszewski, Michał, et al. “High-precision automated reconstruction of neurons with flood-filling networks.” Nature methods 15.8 (2018): 605-610.

Kasthuri, N., Hayworth, K.J., Berger, D.R., Schalek,R.L., Conchello, J.A., Knowles-Barley, S., Lee, D., Va´ zquez-Reina, V., Kaynig, V., Jones, T.R., Roberts, M., Morgan, J.L., Carlos, J.T.,  Seung, H.S., Roncal, W.H., Vogelstein,J.T., Burns, R., Sussman, D.L., Priebe, C.E., ….& Lichtman, J.W. (2015). Saturated reconstruction of a volume of neocortex. Cell, 162(3), 648-661.

Lee, A., Hirabayashi, Y., Kwon, S., Lewis, Jr., T. L., & Polleux, F. (2018). (2018). Emerging roles of mitochondria in synaptic transmission and neurodegeration. Current Opinion in Physiology, 3, 82-93. doi: 10.1016/j.cophys.2018.03.009

Motta, Alessandro, et al. “Dense connectomic reconstruction in layer 4 of the somatosensory cortex.” Science 366.6469 (2019).

Santuy, A., Turegano-Lopez, M., Rodriguez, J. R., Alonso-Nanclares, L., DeFelipe, J., & Merchan-Perez, A. (2018). A quantitative study on the distribution of mitochondria in the neuropil of the juvenile rat somatosensory cortex. Cerebral Cortex, 28, 3673-84. doi: 10.1093/cercor/bhy159

Scheffer, L.K. (2018). Analysis tools for large connectomes. Fronters Neural Circuits, doi: 10.3389/fncir.2018.00085

Sisková, Z., Mahad, D. J., Pudney, C., Campbell, G., Cadogan, M., Asuni, A., O’Connor, V., & Perry, V. H. (2010). Morphological and functional abnormalities in mitochondria associated with synaptic degeneration in prion disease. The American Journal of Pathology, 177(3), 1411–1421. https://doi.org/10.2353/ajpath.2010.091037

Sporns O. (2016). Connectome networks: From cells to systems. In: Kennedy H., Van Essen D., Christen Y. (eds) Micro-, Meso- and Macro-Connectomics of the brain. Research and Perspectives in Neurosciences. Springer, Cham.

Zingg, B., Hintiryan, H., Gou, L., Song, M.Y., Bay, M., Bienkowski, M.S., Foster, N.N., Yamashita, S., Bowman, I., Toga, A.W., & Dong, H. (2015). Neural networks of the mouse neocortex. 156(5), 1096–1111.

 

 

 

 

 

 

 

 

MScBMI Capstone Project

 

Project Title: Morphological Analysis of Mitochondria in Developing     Mouse Cortex

Name(s):

Mentor:

Scientific Advisory Committee:

Project Site:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Personal Statement

 

Presently, advancements in technologies have resulted in the use of large 3D reconstruction in different fields. As students who focus on bioinformatics, we are interested in the practical use of machine learning in real-life health science scenario. Precisely, this research aims at developing a three-dimensional (3D) reconstruction of a 14-days old mouse and finding out the morphological statistics differences of mitochondria between different cortex layers in young mouse brain. The imaging of brain structures is a topic area that has been widely explored. However, there is insufficient research and information on the morphological statistics of mitochondria in the brain of a mouse. Exploring this topic is necessary to give detailed information on the differences in the morphology of these organelles in young mice. This topic is interesting since the morphological statistics of mitochondria is considered a major factor underlying the functional differences. The research will be of great significance to learners, scientists, and researchers. It will have both scientific and academic value. The findings of this research will enable learners to have a deeper understanding of the application of machine learning in the field of biology. Specifically, these learners will comprehensively understand the concept of 3D reconstruction and its applicability to real-life science. Besides, the research will strengthen learners’ skills and understanding of how morphological statistics of mitochondria greatly differ in different cortex layers of young mice. On top of that, learners will have an in-depth understanding of the implications of such differences. Therefore, the expected personal outcome by completing this project to broaden an understanding of the differences of the morphology of mitochondria in mouse and gain necessary knowledge and skills about 3D reconstruction.

Likewise, scientific researchers, especially those who are interested in the mouse brain will benefit from the information provided in this research. The scientists will gain additional insight into the application of a 3D reconstruction in comparing the morphological statistics of mitochondria in young mice cortex layers. There is limited research on this topic, and so this current research will provide valuable information to the existing knowledge. Apart from that, this research will guide scientific researchers who intend to explore a similar topic in the future. Precisely, this research will provide in-depth and valuable information that scientific researchers can rely on to generate reliable and valid evidence concerning the morphological statistics of mitochondria in various brain layers of young mice. Furthermore, scientific researchers will base their study on this research to explore further the role of emerging technologies such as machine learning in improving science. The information on 3D reconstruction will allow scientific researchers to investigate the effectiveness of the method in imaging cells, especially those in remote regions, compared to other tools. Overall, this research will be beneficial in understanding the impact of technological advancements and the rise of machine learning in the field of science. Given the interest of further exploring the use of 3D reconstruction in viewing and understanding brain structure, this project will provide a strong foundation of similar research in the future.

 

Introduction, Background and Significance

 

Brain networks exist only at the cellular level, that is, between particular pre- and post- synaptic neurons that are connected via functionally diverse synapses situation at certain points of cell membranes (Cazemier, Clascá & Tiesinga, 2016). The concept of connectomics refers to an emerging subfield of neuroanatomy that focuses on elucidating brain networks’ wiring with cellular resolution and quantified accuracy (Cazemier, Clascá & Tiesinga, 2016). Therefore, a connectomic analysis should be capable of identifying and measuring the dendrites, axonal path, synapses, and gap junctions of the neurons involved in a specific brain network. Mitochondria are crucial organelles that are responsible for generating high-energy compounds (Sisková et al., 2010). An electronic microscopic 3D method is one of the rising strategies that are employed in identifying certain structures.

The axons and dendrites of nerve cells in the brain of mammals have a high density that light-based imaging methods can hardly resolve the entire network of neuronal wires. Following the recent development in connectomics, advances in 3D electron microscopy allow researchers to have a clear view of each nerve cell and all chemical synapses in a particular piece of brain tissue. Consequently, this increases the possibility of mapping neuronal networks densely and not just sparsely. A connectome can be defined as a connection between regions of the brain of animals (Scheffer, 2018). 3D reconstruction plays a key role in allowing researchers to view images given that it has significantly improved in terms of quality and performance. This imaging tool is useful in the automatic generation of 3D representations of objects, especially those that challenging to model. It speeds up the model generation process for graphics applications. With 3D electron microscopy, it has become possible and easy to image sizeable volumes of neutrophil. An analysis of the anatomical connections of different areas of the mouse visual cortex revealed that there is a division of the mouse cortex into two processing streams (Sporns, 2016).

According to Sporns (2016), involving high-resolution optical imaging and tracking projections across the entire brain of the mouse, has led to another mouse connectome map that charts the directed and weighted anatomical links. Overall, imaging of cells is not an easy task; however, advances in 3D reconstruction has significantly contributed to effective imaging and viewing of various cells, including the mitochondria in the brain. Due to time-consuming manual annotation and deeper development in connectomics, large 3D reconstructions for better analysis are needed. Previous studies have looked into mitochondria in mouse lung tissues or other animals such as fruit flies. As a sample of 14-days old mouse brain cortex is available, we seek to close the research gap. Even though there have been significant advances in the speed of imaging, analyzing 3D image data remains a major challenge. Considering this, dense reconstructions of cortical tissue have been limited to femtoliter-scale volumes; hence, limiting the systematic analysis of axons, neuronal cells, and their dendrites, and the dense connectome between them. Given the 3D complexity and high packing density of many structures, and the fact that axons tend to extend over a long distance to create synapses in remote brain areas, it is challenging to create connectomic maps (Cazemier, Clascá & Tiesinga, 2016). Therefore, developing a 3D reconstruction will be useful in providing an effective connectomic analysis of the mitochondria in different brain layers of young mice.

 

Description of the Problem

 

The structure of brain networks remains largely unknown, and this is attributed to technical difficulties that are involved in imaging and reconstructing the brain in 3D. The imaging of brain structures is associated with various challenges and complexities that require the use of an effective tool. In most cases, a 3D electron microscopy has been widely used by researchers to image nerve cells and synapses in the brain. Although technological advancements have led to the improvement of the performance of a 3D imaging tool, there is a need to explore how 3D can be useful in comparing the differences between brain cells. This method has been extensively used for imaging purposes; however, its effectiveness in the imaging of mitochondria in the brain of mice is not well documented. Therefore, this researcher focuses on investigating the differences in morphological statistics of mitochondria in the brain of young mouse and assessing the effectiveness of characteristic analysis in determining the differences.

In this project, the research will analyze mitochondria’s morphological features based on a dataset acquired in the P14 mouse primary visual cortex. Although several automated methods have been developed because of the requirements of the cell-tracing, those previous work makes us aware of the potential problems such as error rates and massive calculations that need scientists to have further developments. By absorbing thoughts and ways from previous work to build models and reduce merge errors, we may try our best to reduce possible errors while building reconstruction and obtaining results in our research. Our hypothesis are, Shape distribution of Layer 4 mitochondria tend to be more bipolar compared to those of Layer 2/3 mitochondria; Shape distribution of Layer 4 mitochondria tend to be more bipolar compared to those of Layer 2/3 mitochondria..  ​

Furthermore, the researchers are expected to assist with manual annotation, learn to run deep learning models on neural data and perform analysis on objects of biological interest. Particularly, this project will aim at studying a comparison of the morphological statistics of mitochondria in different layers of young mouse brain. Precisely, the researchers will create synaptic annotations and tracing mitochondria by taking advantage of a tolls called “webKnossos”. Apart from the biological importance, their texture creates some difficulty in automatic segmentation. Therefore, we wish to generate a training set from which deep learning models can learn. The researchers can then conduct the comparison of mitochondria from different cortex layers of a young mouse brain. This project is likely to add value to institutions and professionals. The professionals will have the opportunity to gain hands-on experience in data analysis and machine learning. This research aims at enhancing professionals’ understanding of computational biology, develop skillsets for our next career stage, and find some functional differences between brain layers of young mouse. On top of that, institutions such as research centers, information technology organizations, and healthcare facilities will enormously benefit from this research. For instance, the research centers will have access to the additional information provided by this research regarding the connectomic analysis of mitochondria in the brain of young mice. Besides, the research focuses on exploring the practical use of machine learning in the field of biology, particularly covering 3D reconstruction. In this regard, information technology organizations will be well-informed on the effectiveness of the application of 3D reconstruction in health sciences. Consequently, they will have an idea of areas that need improvement and even design new strategies for better imaging of brain structures. Further, the findings of this research provide healthcare institutions with an opportunity to understand the current evidence about the morphological differences of mitochondria in brain layers of young mice. Generally, this project is more likely to be important to different stakeholders.

 

Literature Review

 

While extant studies examine the neuronal exchanges in the brain cortex, the broader organization of the communication network across the cortex remains unclear. Yet, these networks underlie the expression of cognition and behavior. This expression requires an integration of multiple sensory inputs, harmonization of several motor outputs, and synchronization of activities within the system that connect the two (Zing et al. 2015). Consequently, researchers must develop a brain-wide connectivity map for all regions of the gray matter, which will enhance understanding of the neural influences on behavior and cognitive function.

Several automated techniques have been used to probe the structure of the neuronal cortex. Firstly, Kasthuri et al. (2015) explored the structure of the neural tissue using automated techniques. The researchers opined that reconstruction effort of the mammalian brain is a costly undertaking, and substantial savings would be produced by examining overlaps of dendrites and axons to infer connectivity. However, reconstruction of an entire mammalian brain would result in a large and superfluous dataset, which would present significant challenges in processing. Consequently, Kashturi et al. used a small piece (1,500 µm3) of neocortical tissue to reconstruct all the connectivity within the cortex. Using saturated reconstruction, the researchers identified each point of contact between excitatory axons and dendritic spines. They found that nine different axons appose each spine, and only a single axon institutes a synapse. Kasthuri et al. offer sufficient explanation of the complex nature of mapping the neuronal cortex. Nonetheless, the research provides a simplified illustration of mapping the brain cortex and can be replicated on a larger scale.

Changes in neuronal connections influence experiential learning. A study by Dorkenwald et al (2019) presented a large map of connectivity between cortical neurons using electron microscopy, which demonstrated the learning processes used by the cortex. These researchers demonstrate that synapse size can be modelled using a binary and analog variable derived from a log-normal distribution. Using data from a mouse’s primary visual cortex, the research finds that synapse size equates to a binary variable combined with a log-normal continuous variable. Further, the research finds that synaptic plasticity-or the biological processes that underlie learning and memory- modifies these binary and log-normal variables. \

Similarly, Motta, Alessandro, et al. (2019) raised the efficiency of these methods by improving the automated segmentation quality, analyzing the automated segmentation for locations of likely errors and directing the human work to these locations and optimizing human data interaction by helping annotators to immediately understand the problem to be solved.

Mitochondria are important organelles for brain cells as they facilitate energy production in the form of ATP, thus allowing the brain neurons to send messages and coordinate body functions via synaptic transmission, calcium uptake, which allows electrical signals to be transmitted between nerve synapses, and the catabolism of reactive oxygen species (ROS), which are harmful to brain cells. Hence, the level of activity in a brain cortex determines the number of mitochondria found therein and their size. They are transported from one region to another via the axons and dendrites of nerve cells. The mouse cerebral cortex is made up of six layers with a uniform laminar structure. Layers 1, 2/3, and 4 constitute the upper layers and process information locally, which is attributed to their localized intracortical connections, while layers 5, 6, and 6b form the deep layers with longer connections to subcortical targets, such as the spinal cord and thalamus, among others, and the opposite hemisphere (Belgard et al., 2011). The number of mitochondria in the upper layers of the mouse brain cortex increases progressively from layer 1 to layer 4, with the highest volumes located in the dendrites of layer 2/3 to facilitate the intracortical connections between the layers. However, most neuroscientists use rats as it is difficult to analyze the mouse brain cortex accurately due to the small amount of tissue available.

In their study of the mitochondria found in the somatosensory cortex of the juvenile rat and the distribution of the organelles between nonsynaptic, axonal, and dendritic processes, Santuy and colleagues (2018) found out that the volume fraction of mitochondria in layer 4 was 7.74% in comparison to the 4.59% observed in layer 1, with statistically significant disparities between layers 4 and 2/3 and all the other layers. The mitochondrial volume was also highest in the dendrites of layer 3 at 48.70%, compared to layer 2 at 46.50% and layer 4 at 43.92%, and at the excitatory axons of layer 3 (14.35%), relative to layer 4 at 13.53% (Santuy, Turegano-Lopez, Rodriguez, Alonso-Nanclares, DeFelipe, & Merchan-Perez, 2018). These differences show that the highest energy demand levels for dendrites are at the connection between layer 2/3 and 4, primarily because layer 2/3 is at the middle of the upper cortex. However, considering all the mitochondria in the dendrites, excitatory and inhibitory axons, and nonsynaptic elements, the volume fraction of the organelles was highest in layer 4 at 7.74%, compared to layer 3 at 6.40% and layer 2 at 5.22% (Santuy et al., 2018). Therefore, it can be deduced that there is a positive correlation between mitochondrial volume fraction and synaptic density and activity. Layer 4 of the cerebral cortex has the most mitochondria because it receives long-distance thalamic afferences with high firing rates to facilitate efficient transmission of signals (Santuy et al., 2018). The mitochondria in the dendrites also had distinct morphologies to those found in the axons. Dendritic mitochondria were in high volumes and with long, tubular shapes due to the high degree of fusion in the region, whereas those in the axons were small and punctate due to their high levels of fission (Lee, Hirabayashi, Kwon, Lewis, Jr., & Polleux, 2018). This unique morphology influences the function of the mitochondria at vesicle release sites as the long, tubular shapes promote increased ATP synthesis to pump sodium ions out of cells as the post-synaptic side of layer 4 is more concentrated with the ions compared to the presynaptic side.

Brandt et al.’s (2017) study finds that age-related changes in mice mitochondria were more pronounced in heart and liver cells. The researchers explore the impact of ageing on mitochondrial function and ultrastructure in fruit flies and mice. They find that the mitochondrial ultrastructure and function remain unchanged in heart cells, while liver and kidney cells show age-related alterations in membrane morphology. Consequently, the study establishes a clear link between functional decline of mitochondria and the architecture of the inner membrane. Further, they observe that these changes depend on the organism, as well as the extent to which the tissues are protected against damage. The study offers important insights on the effects of ageing on animals with different lifespans.

Brandt et al.’s study used a more representative sample of mice and fruit flies. The sampled mice ranged from 20 to 96 weeks of age. The study estimated the average lifespan of the mice as 107 weeks for males and 114 for females. Further, the study used high-resolution respirometry and electron cryo-tomography to assess mitochondrial function, thereby enhancing the accuracy of measurement. Consequently, the researchers identified abnormal morphologies in mitochondria including widened crystae, crystae membrane swirls, voids, granules, and membrane enclosures (Brandt et al., 2017). However, the research did not examine brain cells, which play an important role in the biogenesis of the mitochondria. Nonetheless, the expanded age-range of the sample offered an expanded view of the changes in morphology over mice’s lifetime.

Developing a full-scale map of the brain-wide neuronal circuitry presents major challenges for researchers. The reconstruction effort is complex and costly and results in large datasets that add to the complexity. Yet, understanding the structure and connections of the brain helps in analyzing the cognitive and behavioral characteristics of the subjects. Recent developments in imaging technologies and algorithmic processing have improved the reconstruction modelling capacity, but major hurdles remain in integrating large data sets to develop complete images. While the technologies are still evolving, major hurdles remain in developing full-scale brain models needed to study the nature and size of synapses, and their influence on cognition and behavior. The reviewed studies reflect a gradual development in the right direction but leave significant research and practice gap. Therefore, this research seeks to close this evident research gap.

Description of the project and anticipated results

 

This section describes the methodology and timeline of the project, from inception to completion. The major steps of the project include:

  1. Reading papers and get knowledge of previous work
  2. Using WebKnossos to trace mitochondria in samples of different parts of young mouse brain
  3. Validating manual tracing accuracy
  4. Using Unet to finish tracing automatically the complete datasets from layer 4 and layer 2/3
  5. Analyzing the morphological statistics differences of mitochondria from two layers.

 

Various methods can be used in tracing mitochondria, including webKnossos which is an automated method, manual tracing, and UNet. In this study, the researcher intends to use the automated methodology, WebKnossos to significantly trace mitochondria. Tracing and analyzing as the major steps in this study that requires much attention. The major risk associated with this study that is worth addressing is that there may be some errors during manual tracing of mitochondria. Thus, in order to validate manual tracing accuracy, two ways of quality testing are desired. First, the tracing task will be divided and traced by two students individually. Thus, works can be examined by each other. Second, once manual trace is done, the piece will be divided into two potions, training set and testing test. Machine learning will be implemented to validate accuracy. Once manual tracing is done, it can be uploaded to UNet, and used as training dataset for future automated tracing.

Additionally, analysis is essential and should provide necessary information regarding the main contribution of the project. The phase of analyzing image data will be mainly completed using python. Possibly, the targeted measurements may include the 3D shape, volume size, spatial distribution, etc. With that, further analysis will be based on comparison with mitochondria in different mouse brain layers. The researchers will be using Jupyter Notebook, a python shell, and two handy python packages, numpy and ndimage. The idea is to transfer image data to digital data for numerical analysis.

The problems, especially merge errors that may occur resulting from the automated tracing of mitochondria is related to fragmentation of the organelle which is an indication of unhealthy cells and is caused by high scanning laser power, and high energy and low wavelength laser. Furthermore, damage to the tissue sample could lead to break of cell membrane, and carbon dust could result in the appearance large cluster of black matter. Another problem is immotile mitochondria resulting from high speed of imaging. These factors interfere with machine tracing; hence, affecting the overall outcome. Afterall, p14 mouse brain tissue is fragile. However, with proper manual tracing, machine can be trained to avoid the error. The equipment or software needed to conduct the study is laboratory computers, particularly, a supercomputer is necessary. When it comes to costs, there is no need for additional costs.

The anticipated results of the morphological analysis of mitochondria will be presented in diagrams and tables of analytic results. The outcome will be based on comparing the morphological statistics of the mitochondria from different brain layers. Since the brain is highly sensitive to the morphological changes caused by aging, it is possible that the synaptic mitochondria of young mouse won’t show much difference due to its young age. Nevertheless, our project results will contribute to Kasthrui’s lab, as well as to the field of connectomics.  Overall, the anticipated outcome of the analysis will be presented in tables, or diagrams appropriately.

The project’s benefits to the sponsor organization or research field is that manual tracing will assist in alleviating some of the problems associated with machine learning workflow. Likewise, the morphological analysis of mitochondria will enable individuals to gain more biological outcomes from the datasets. Therefore, this project will be beneficial to both researchers and sponsor organizations.

 

Project Methodology — Tracing

The purpose of tracing is to identify and locate mitochondria in brain cells. This step is fundamental since only accurate tracing could lead to significant results, otherwise the results would just be meaningless. This section shows the tracing phase of our project in detail.

The image above is the process of manual tracing, we identify mitochondria with eyes and paint them with different color. This process can be very tedious and time consuming since there are thousands of mitochondria in our dataset, so we would hope machine can do the tracing work for us. However, since the field of connectomics is relatively new and suffers from shortage of high-quality training data, automated tracing could be problematic at this point. Therefore, we traced manually a portion of the data to create training sets to bootstrap different machine learning models. We cut roughly 3 percent of each complete data as our sample dataset and assist those with manual annotation. The final number of mitochondria we traced in sample dataset manually are layer 4 with 288 mitochondria, and layer 2/3 with 200 mitochondria. Then we used the manual traced results as train set for their complete layer dataset separately, sample layer 4 for the complete layer 4 dataset, while sample layer 2/3 for the complete layer 2/3 dataset. After a test run, we noticed some training error shown below.

 

Training errors are identified at the bottom left as well as the top right corner along the edges of cell membrane. Machine falsefully labeled some areas as mitochondria. This can be caused by color similarity which can be misleading to the machine. We manually fixed those training error before implementing the algorithm on the complete datasets. ​ After all, our complete annotated layer datasets are:

Layer 4: five gigabytes in file size with 6948 traced mitochondria

layer 2/3: five gigabytes in file size with 6837 traced mitochondria

 

Project Methodology – Analysis

 

To study features of mitochondria of biological interests, we converted brain cell image data into the forms of matrices for numerical analyses. This section explains the methodology for each steps of the analysis. For a clear illustration purpose, the section uses the sample dataset, sample set from layer 4 with 288 mitochondria.

The first step was to study mitochondria’s size, we calculated the mean and the standard. The calculated for mean for layer 4 sample mitochondria was 44257 while the standard deviation was 49590. The mean and the standard deviation would be essential for comparing of the results to determine if the two layers are comparable. To ensure that we got the correct results from the comparison of the mean and the standard deviation, we used histograms as shown in the figures below.

 

Our interest was also in determining the length and the direction of the mitochondrion, in order to do that, we performed PCA, principal component analysis, on each mitochondrion to get three principal component and its growing direction. The plot below is the result of sample layer 4, each cone shape is a representation of mitochondria.

 

However, since the axis is normalized in terms of length during PCA process, we decided to transform back the axis to their physical coordinates. After that, we got the length of the three essentials axis for each mitochondrion. Those axis are similar to the length, width, and height of a cube. Nevertheless, since mitochondria are ellipsoid in shape, we got one long axis and two short axis. We defined the long axis as its pointing and growing axis, because in our physical interpretation, it would seem like the mitochondrion is pointing at its foci; and when viewed from a mathematical dimension, the long axis outgrows the other two axis. Below are the length distribution graph of one growing axis and two short axis.

Growing axis:

Short Axis 1:

Short Axis 2:

 

The three axis were then combined in one graph for the difference to be studied and later we combined the three axis into one distribution graph to ensure our findings are correct.

Another strategy that we considered, and we carried out was on k-means clustering based on direction. This was to help us determine the direction that each mitochondrion took during their growth. We agreed that each arrow would be representation of a mitochondria and the clustering was according to their orientation. We faced a slight challenge in determining the head and the tail of foci since the shape of the mitochondrion is ellipsoid. However, this did not hinder us from continuing with our experiment. Since we never wished to stop halfway, we considered conversion of each mitochondrion dimension coordinates to ensure that it is possible to compare. This made it possible to determine the foci according to the direction that each took. We classified the data points into five groups. Each color represents a group, and each arrow in one group points approximately the same direction. This is clearly shown in the diagram below.

 

The number of mitochondria in each cluster: [113 77 89 67 59].

 

At the end, we performed length distribution of growing axis based on k-means clustering. For us to succeed, we considered using the results we had obtained from k-means clustering. This would help us in studying the growing patterns of the mitochondrion.

 

 

The length distribution of growing axis based on k-means clustering result with distribution curve is as follows:

 

Preliminary Findings and Results

 

After implementing codes on the completed datasets, the results are shown and compared in this section. For size distribution of mitochondria from the complete layer 4 set, the mean is 18345, and standard deviation is 34965. While size distribution for the complete layer 2/3 set, mean is 16597 with standard deviation of 24730. As shown in the graph, the graph does not start exactly on zero on the x axis. This is because machine labeled a fair amount of dust segment as mitochondria, therefore, we set the minimum threshold as 2000 nanometer.

 

Size distribution

Layer 4

(Mean: 18345, std: 34965)

 

Layer 2/3

(Mean: 16597. Std: 24730)

 

Comparison result: As shown in the histograms, the mean from two size distribution graph is similar, both have very large standard deviation meaning the points are far away from the mean. Nevertheless, standard deviation from layer 4 is greater. This is because layer 4 contains more extremely small mitochondria, as well as more extremely big mitochondria. While layer 2/3 mitochondria tend to cluster at a lower middle size. However, the total number of those extreme ones in layer 4 are not significant enough to change the bell curve. From the graph, the bell curve looks similar, both extremely right skewed, meaning the majority of mitochondria from both datasets are small ones. The extremely big ones from layer 4 makes it curve move slightly a bit towards right.  So, we cannot conclude that size of mitochondria is brain layer 4 are more bipolar distributed compared to those from layer 2/3. Overall, the size distribution is similar.

 

PCA analysis

Layer 4:

 

Layer 2/3

Comparison result: It is difficult to find difference based on visual comparison since the dataset is huge. At this point, they look similar, we will quantify the characteristics for further analysis.

 

Length Distribution of Growing/point axis

Layer 4

Layer 2/3

 

Length Distribution Short Axis 1

Layer 4

Layer 2/3

 

Length Distribution of Short Axis 2

Layer 4

Layer 2/3

Combined Length Distribution of Three Axis

Layer 4 

Layer 2/3

Comparison result: Both mitochondria from brain layer 4 and layer 2/3 displaced a bell curve of gaussian distribution. However, layer 4 has shown numbers that are larger in the two short axis, and smaller in the growing and pointing axis compared to those form layer 2/3. This means that mitochondria from layer 4 tend to be more sphere-shaped, while those from layer 3 tend to be ellipsoid-shaped. Nevertheless, the difference is not much significant.

 

K-means Clustering Based on Orientation

Layer 4

 

Number of mitochondria in each cluster: [1515 1457 1347 1365 1264]

 

Layer 2/3

 

Number of mitochondria in each cluster: [1235 1435 1362 1179 1626]

Five Approximate Direction in 2D

 

Comparison result: The number of mitochondria in each cluster from the two brain layers are similar. Each cluster has more than 1200, and less than 1600 mitochondria. This means, although there are certain directions that have slightly more mitochondria pointing, there exists no dominant direction that all mitochondria tend to point at. We performed a length distribution of growing axis based on k-means clustering results in order to study patterns in growing direction.

 

Length Distribution of Growing Axis Based on K-means Clustering Result

Layer 4

Layer 2/3

 

Comparison result: Overall, mitochondria from the two brain layers have very similar patterns in orientation. From the graph, for both brain layers, mitochondria with long growing axis tend to point in the purple direction. Mitochondria with short growing axis tend to point to the green direction. The length of growing axis from layer 4 mitochondria is comparable with the length of those from layer 2/3 in both the green and purple direction. This represents that mitochondria pointing to the purple direction are more likely to outgrow those pointing in the green direction. The difference is that, in the orange and red direction, layer 4 contains larger mitochondria compared to layer 2/3. While in the blue direction, layer 2/3 consists of larger mitochondria compared to layer 4.

 

Conclusion and limitation

 

Overall, morphological statistics of mitochondria between brain cortex layer 4 and brain cortex layer 2/3 of developing mouse exist no significant differences. However, few minor variations can be identified based on our study. First of all, layer 4 contain slightly more mitochondria, and their size tend to be more bipolar compared to those of layer 2/3, namely extremely large and extremely small ones. Mitochondria from layer 4 tend to be more sphere-shaped while those from layer 2/3 tend to be more long-ellipsoid-shaped. Although there is no dominant direction that mitochondria are pointing at for each layer, there are small differences in size between each clustering direction.

The only regret left from this project is that we did not confirm any of our hypothesis, since there exists no significant morphological difference. This could because that the mouse is too young to show significant difference in the brain layer, or simply mitochondria are similar between the two brain layers. We are positive that more information can be inferred if compared to the same layer region of an adult mouse brain. Nevertheless, this project is significant in the way that with the same methodology and the same codes, we can analyze and compare morphological difference of mitochondria from any parts of any mammal, if we have the specific data set available. For instance, if one wishes to study the morphological difference of mitochondria between dog’s lung tissue and its brain tissue, he can simply go through the manual tracing process, implement the same machine algorithm, and change the data input in the coding. Moreover, after accumulation of accurate training tracing data, the stage of manual tracing will no longer be necessary. Much more information can be drawn, and further studies can be conducted based on this project.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

References

Belgard, T. G., Marques, A. C., Oliver, P. L., Abaan, H. O., Sirey, T. M., Hoerder-Suabediseen, A., … Ponting, C. P. (2011). A transcriptomic atlas of mouse neorcortical layers. Neuron, 71(4), 605-616. doi: 10.1016/j.neuron.2011.06.039

Brandt, T., Mourier, A., Tain, L.S., Partridge, L., Larsson, N., & Kühlbrandt, W. (2017). Changes of mitochondrial ultrastructure and function during ageing in mice and Drosophila. eLife Journal, 6, 1-19.

Cazemier, J. L., Clascá, F., & Tiesinga, P. H. (2016). Connectomic analysis of brain networks: Novel techniques and future directions. Frontiers in Neuroanatomy, 10, 110. https://doi.org/10.3389/fnana.2016.00110

El-Merhie, N., Baumgart-Vogt, N., Pilatz, A., Pfreimer, S., Pfeiffer, B., Pak, O., Kosanovic, D., Seimetz, M., Schermuly, R.T., Weissmann, N., Karnati, S. (2017). Differential alterations of the mitochondrial morphology and respiratory chain complexes during postnatal development of the mouse lung. Oxidative Medicine and Cellular Longevity, 1, 1-22.

Januszewski, Michał, et al. “High-precision automated reconstruction of neurons with flood-filling networks.” Nature methods 15.8 (2018): 605-610.

Kasthuri, N., Hayworth, K.J., Berger, D.R., Schalek,R.L., Conchello, J.A., Knowles-Barley, S., Lee, D., Va´ zquez-Reina, V., Kaynig, V., Jones, T.R., Roberts, M., Morgan, J.L., Carlos, J.T.,  Seung, H.S., Roncal, W.H., Vogelstein,J.T., Burns, R., Sussman, D.L., Priebe, C.E., ….& Lichtman, J.W. (2015). Saturated reconstruction of a volume of neocortex. Cell, 162(3), 648-661.

Lee, A., Hirabayashi, Y., Kwon, S., Lewis, Jr., T. L., & Polleux, F. (2018). (2018). Emerging roles of mitochondria in synaptic transmission and neurodegeration. Current Opinion in Physiology, 3, 82-93. doi: 10.1016/j.cophys.2018.03.009

Motta, Alessandro, et al. “Dense connectomic reconstruction in layer 4 of the somatosensory cortex.” Science 366.6469 (2019).

Santuy, A., Turegano-Lopez, M., Rodriguez, J. R., Alonso-Nanclares, L., DeFelipe, J., & Merchan-Perez, A. (2018). A quantitative study on the distribution of mitochondria in the neuropil of the juvenile rat somatosensory cortex. Cerebral Cortex, 28, 3673-84. doi: 10.1093/cercor/bhy159

Scheffer, L.K. (2018). Analysis tools for large connectomes. Fronters Neural Circuits, doi: 10.3389/fncir.2018.00085

Sisková, Z., Mahad, D. J., Pudney, C., Campbell, G., Cadogan, M., Asuni, A., O’Connor, V., & Perry, V. H. (2010). Morphological and functional abnormalities in mitochondria associated with synaptic degeneration in prion disease. The American Journal of Pathology, 177(3), 1411–1421. https://doi.org/10.2353/ajpath.2010.091037

Sporns O. (2016). Connectome networks: From cells to systems. In: Kennedy H., Van Essen D., Christen Y. (eds) Micro-, Meso- and Macro-Connectomics of the brain. Research and Perspectives in Neurosciences. Springer, Cham.

Zingg, B., Hintiryan, H., Gou, L., Song, M.Y., Bay, M., Bienkowski, M.S., Foster, N.N., Yamashita, S., Bowman, I., Toga, A.W., & Dong, H. (2015). Neural networks of the mouse neocortex. 156(5), 1096–1111.

Quick Homework Essays
Calculate your paper price
Pages (550 words)
Approximate price: -

Why Work with Us

Top Quality and Well-Researched Papers

We always make sure that writers follow all your instructions precisely. You can choose your academic level: high school, college/university or professional, and we will assign a writer who has a respective degree.

Professional and Experienced Academic Writers

We have a team of professional writers with experience in academic and business writing. Many are native speakers and able to perform any task for which you need help.

Free Unlimited Revisions

If you think we missed something, send your order for a free revision. You have 10 days to submit the order for review after you have received the final document. You can do this yourself after logging into your personal account or by contacting our support.

Prompt Delivery and 100% Money-Back-Guarantee

All papers are always delivered on time. In case we need more time to master your paper, we may contact you regarding the deadline extension. In case you cannot provide us with more time, a 100% refund is guaranteed.

Original & Confidential

We use several writing tools checks to ensure that all documents you receive are free from plagiarism. Our editors carefully review all quotations in the text. We also promise maximum confidentiality in all of our services.

24/7 Customer Support

Our support agents are available 24 hours a day 7 days a week and committed to providing you with the best customer experience. Get in touch whenever you need any assistance.

Try it now!

Calculate the price of your order

Total price:
$0.00

How it works?

Follow these simple steps to get your paper done

Place your order

Fill in the order form and provide all details of your assignment.

Proceed with the payment

Choose the payment system that suits you most.

Receive the final file

Once your paper is ready, we will email it to you.

Our Services

No need to work on your paper at night. Sleep tight, we will cover your back. We offer all kinds of writing services.

Essays

Essay Writing Service

No matter what kind of academic paper you need and how urgent you need it, you are welcome to choose your academic level and the type of your paper at an affordable price. We take care of all your paper needs and give a 24/7 customer care support system.

Admissions

Admission Essays & Business Writing Help

An admission essay is an essay or other written statement by a candidate, often a potential student enrolling in a college, university, or graduate school. You can be rest assurred that through our service we will write the best admission essay for you.

Reviews

Editing Support

Our academic writers and editors make the necessary changes to your paper so that it is polished. We also format your document by correctly quoting the sources and creating reference lists in the formats APA, Harvard, MLA, Chicago / Turabian.

Reviews

Revision Support

If you think your paper could be improved, you can request a review. In this case, your paper will be checked by the writer or assigned to an editor. You can use this option as many times as you see fit. This is free because we want you to be completely satisfied with the service offered.