Back

DNA shape complements sequence-based representations of transcription factor binding sites

DeFord, P. M.; Taylor, J.

2019-06-17 bioinformatics
10.1101/666735 bioRxiv
Show abstract

The position weight matrix (PWM) has long been a useful tool for describing variation in the composition of regions of DNA such as transcription factor (TF) binding sites. It is difficult, however, to relate the sequence-based representation of a DNA motif to the biological features of the interaction of a TF with its binding site. Here we present an alternative strategy for representing DNA motifs - called Structural Motif (StruM) - that can easily represent different sets of structural features. Structural features are inferred from dinucleotide properties listed in the Dinucleotide Property Database. StruMs are able to specifically model TF binding sites, using an encoding strategy that is distinct from sequence-based models. This difference in encoding strategies makes StruMs complementary to sequence-based methods of TF binding site identification.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.