Student Work at Broad Institute
Sreshtaa Rajesh, Class of 2019, recently completed a program in the computational epigenomics lab at Broad Institute. As she reports, “We were focused on histones–proteins that DNA is wound around. Changes to the chemical composition of these histones can change the accessibility of the DNA around it, altering gene expression. The data that we got aimed to map the frequency of certain histone modifications along the genome, however we were only interested in the regions that had elevated levels. Over the last year, my lab had developed an algorithm that aimed to parse through the data and differentiate background (the normal distribution of the data) from the signal (the areas with elevated levels of histone modifications, therefore the areas that were of interest).
The algorithm was tested and validated on Broad-Institute generated data, but before it could be implemented and introduced to the scientific community, it needed to be validated on non-Broad generated data (because each lab will introduce its own bias into data). I imported non-Broad datasets, adapted the algorithm for use on them, and performed visual and quantitative analyses. Statistical results between each labs’ data came back insignificant, suggesting that the algorithm does not perform significantly differently when introduced to different biases. These results are promising in not only validating the algorithm, but for developing a better fit-quality metric that can give us an accurate, quantitative way of analyzing how well the algorithm works on future datasets.”