The prediction of transcription factor binding sequences in DNA is an important problem in computational biology with implications for many biological phenomena. Current approaches have used ChIP-seq data to learn a representation of a given transcription factors binding affinity. One such method is the k-mer motif and alignment clustering (KMAC) algorithm to produce k-mer set motif (KSM) representations. However, the use of ChIP-seq data is relatively expensive because a new dataset has to be derived for each transcription factor. In order to improve this, we propose to adapt the KMAC/KSM approach to use DNase-seq data, which can be treated as multiplexed ChIP-seq data, in order to produce binding site representations for multiple transcription factors from a single DNase-seq experiment.

I am participating in SuperUROP to get more hands-on experience in a research setting. I am a course 6 student but have also done a lot of work related to biology, including many terms doing wet lab research. This computational biology project looks like an exciting way to marry these two interests.