Leon Peshkin and some high-end software projects
I sat with Leon for a while in what they call the 'interactive room', a lounge area in the center of the department where faculty, staff and students can mingle. Leon's a computer scientist and he seems to be free roaming--not attached to any particular lab and working on several projects at once.
One project is to use a computer to find and identify cells that are dividing abnormally. To determine what genes affect cell division, the folks here developed thousands of cell types, each with a single gene knocked out. They then grow the groups of these cells on plates and look to see if any of them divide abnormally. The idea is that if a cell type divides abnormally, then the gene you knocked out must somehow be involved with cell division. Crude but effective.
Rather than watch each group of cells divide, an impossible task, they take pictures of the plates at repeated intervals as the cells grow and divide. They then look through the pictures counting abnormal cell divisions. Now, even in a normal group of cells they'll get some abnormal divisions, so the key is to look for enough abnormalities that they know you're on to something. In math-speak they're looking for something statistically significant.
The problem is that there are thousands of pictures to look at, and each has hundreds of cells. Again a very daunting task! So the folks here turn to computers to do the counting for them. Leon is working on algorithms to find the cells in each image, determine whether they are dividing, and then whether they are dividing abnormally. Very tricky because cells don't have very clear boundaries.
Another project Leon is working on is linguistics of protein structures. The idea is to look at the structure of proteins as a language with particular semantic rules and then using linguistic theories to find patterns. Cool stuff.
One project is to use a computer to find and identify cells that are dividing abnormally. To determine what genes affect cell division, the folks here developed thousands of cell types, each with a single gene knocked out. They then grow the groups of these cells on plates and look to see if any of them divide abnormally. The idea is that if a cell type divides abnormally, then the gene you knocked out must somehow be involved with cell division. Crude but effective.
Rather than watch each group of cells divide, an impossible task, they take pictures of the plates at repeated intervals as the cells grow and divide. They then look through the pictures counting abnormal cell divisions. Now, even in a normal group of cells they'll get some abnormal divisions, so the key is to look for enough abnormalities that they know you're on to something. In math-speak they're looking for something statistically significant.
The problem is that there are thousands of pictures to look at, and each has hundreds of cells. Again a very daunting task! So the folks here turn to computers to do the counting for them. Leon is working on algorithms to find the cells in each image, determine whether they are dividing, and then whether they are dividing abnormally. Very tricky because cells don't have very clear boundaries.
Another project Leon is working on is linguistics of protein structures. The idea is to look at the structure of proteins as a language with particular semantic rules and then using linguistic theories to find patterns. Cool stuff.
0 Comments:
Post a Comment
<< Home