Memory Matching Networks for Genomic Sequence Classification

Jack Lanchantin, Ritambhara Singh, Yanjun Qi

Abstract: When analyzing the genome, researchers have discovered that proteins bind to DNA based on certain patterns on the DNA sequence known as "motifs". However, it is difficult to manually construct motifs for protein binding location prediction due to their complexity. Recently, external learned memory models have proven to be effective methods for reasoning over inputs and supporting sets. In this work, we present memory matching networks (MMN) for classifying DNA sequences as protein binding sites. Our model learns a memory bank of encoded motifs, which are dynamic memory modules, and then matches a new test sequence to each of the motifs to classify the sequence as a binding or non-binding site.

TL;DR: We extend the Matching Network Model of Vinyals et al. (2016) for a general classification setting, where a learned memory is useful.

Conflicts: cs.virginia.edu

Keywords: Deep learning, Supervised Learning, Applications