Team:Berkeley/Project/Automation

From 2012.igem.org

(Difference between revisions)
Line 58: Line 58:
<div class="col1" align="justify">
<div class="col1" align="justify">
<p>
<p>
-
We can accommodate libraries of up to one million members through our MiCodes design scheme. With a data set this large, we needed to develop methods to make MiCodes a high-throughput screening technique. We realized that two aspects of microscopy needed to be automatable: image acquisition and image processing.  
+
We can accommodate libraries of over a million members through our MiCodes design scheme. With a data set this large, we needed to develop methods to make MiCodes a high-throughput screening technique. We realized that two aspects of microscopy needed to be automatable: image acquisition and image processing.  
</p>
</p>
</div>
</div>
Line 67: Line 67:
     <img src="https://static.igem.org/mediawiki/2012/4/42/Project3_subheader1.jpg" width="980">
     <img src="https://static.igem.org/mediawiki/2012/4/42/Project3_subheader1.jpg" width="980">
<p>
<p>
-
If a MiCode library size is very large, it is important to find a time-efficient method of taking many pictures of yeast cells. Because over the summer we worked on small scale MiCode libraries, all of our images were taken with a regular fluorescence microscope. Nevertheless, automating the image acquisition of MiCodes is definitely achievable with different imaging tools. There are many possible imaging technologies to choose from, as long as there is a fluorescence lamp to excite the fluorescent proteins of the MiCode. As an example, we could simply use a fluorescence microscope with an automated stage to speed up image acquisition. We could also use a confocal microscope to help us identify organelles, since it has the capability to take layers of images (through a thick cell) and compile the images  into a clearer resulting image. Another technology we considered was flow cytometry (such as FACS), in which a fluid stream of cells is shot through a fluorescent laser beam to excite fluorescence. However, flow cytometers for the most part can only detect fluorescence on the outer surface of the cell. This means it cannot detect the MiCode fluorescence on the organelles inside the cell, and cannot allow the user to observe any other phenotypes aside from from the outer fluorescence.
+
If a MiCode library size is very large, it is important to find a time-efficient method of taking many pictures of yeast cells. This summer we dealt with libraries on the order of 10^3 members, and our images were taken with a regular fluorescence microscope. Larger libraries would most likely make use of automated stages to speed up image acquisition.
</p>
</p>
</div>
</div>
Line 124: Line 124:
<br><br>
<br><br>
-
We applied machine learning concepts in order to tackle automation of MiCode identification. The following illustrates general workflow principles for machine learning:<br>
 
-
1. Design of features associated with organelles, using image analysis software. <br>
 
-
2. Iterative Machine learning -> computer presents small sample of cells to biologist, who batch scores and returns categorized cells.<br>
 
-
3. Computer scores phenotypes among the sample and refines rules by comparing to biologist's scores.<br>
 
-
4. When accuracy of identification through the rules is sufficient, classification is automatically run on all cells in experiment.<br>
 
-
 
-
 
-
<br><br>
 
-
 
-
 
We quickly realized that the most essential portion of automation was determining features. In the context of organelle detection, we developed pipelines to collect geometric measurements characteristic of each organelle (actin, cell periphery, vacuolar membrane, and nucleus). <br> <br>
We quickly realized that the most essential portion of automation was determining features. In the context of organelle detection, we developed pipelines to collect geometric measurements characteristic of each organelle (actin, cell periphery, vacuolar membrane, and nucleus). <br> <br>
 +
To acheive this, we imaged a small sample size of about 100 cells with varying phenotypes, and determined feature sets associated with each of the four organelles. We used CellProfiler, an open-source software designed for quantitative analysis of cellular phenotypes. The software is available at their <a href="http://www.cellprofiler.org/">website</a>. We used this software to build pipelines that demonstrate the feasibility of automated identification of cellular phenotypes. Our software proved capable of successfully identifying MiCoded nuclei 95% of the time. In the future, with a large, already-constructed dataset, we would likely employ machine learning techniques to improve the automated identification.<br> <br>
-
To acheive this, we imaged a small sample size of about 100 cells with varying phenotypes, and determined feature sets associated with each of the four organelles. We used CellProfiler, an open-source software designed for quantitative analysis of cellular phenotypes. The software is available at their <a href="http://www.cellprofiler.org/">website</a>. We used this software to build pipelines that demonstrate the feasibility of applying machine learning for identify cellular phenotypes. <br> <br>
+
<br> <br>
-
 
+
-
We developed the substantial part of the workflow, the feature design, to demonstrate the feasibility of automated MiCode identification. Because we spent the summer constructing the MiCodes, we did not have sufficient time to build a comprehensive training set to process and refine the rules and algorithms for scoring phenotypes.<br> <br>
+

Revision as of 19:31, 2 October 2012

header
iGEM Berkeley iGEMBerkeley iGEMBerkeley

Mercury

We can accommodate libraries of over a million members through our MiCodes design scheme. With a data set this large, we needed to develop methods to make MiCodes a high-throughput screening technique. We realized that two aspects of microscopy needed to be automatable: image acquisition and image processing.

If a MiCode library size is very large, it is important to find a time-efficient method of taking many pictures of yeast cells. This summer we dealt with libraries on the order of 10^3 members, and our images were taken with a regular fluorescence microscope. Larger libraries would most likely make use of automated stages to speed up image acquisition.


How the cell segmentation was done in Matlab.


How we created pipelines in Cell Profiler to detect organelles.

We quickly realized that the most essential portion of automation was determining features. In the context of organelle detection, we developed pipelines to collect geometric measurements characteristic of each organelle (actin, cell periphery, vacuolar membrane, and nucleus).

To acheive this, we imaged a small sample size of about 100 cells with varying phenotypes, and determined feature sets associated with each of the four organelles. We used CellProfiler, an open-source software designed for quantitative analysis of cellular phenotypes. The software is available at their website. We used this software to build pipelines that demonstrate the feasibility of automated identification of cellular phenotypes. Our software proved capable of successfully identifying MiCoded nuclei 95% of the time. In the future, with a large, already-constructed dataset, we would likely employ machine learning techniques to improve the automated identification.




How and why we checked our library.

Retrieved from "http://2012.igem.org/Team:Berkeley/Project/Automation"