Prototype Matching Networks for Large-Scale Multi-Label Classification

Jack Lanchantin, Arshdeep Sekhon, Ritambhara Singh, Yanjun Qi

Feb 12, 2018 (modified: Feb 20, 2018) ICLR 2018 Workshop Submission readers: everyone
  • Abstract: One of the fundamental tasks in understanding genomics is the problem of predicting Transcription Factor Binding Sites (TFBSs). With more than hundreds of Transcription Factors (TFs) as labels, genomic-sequence based TFBS prediction is a challenging multi-label classification task. There are two major biological mechanisms for TF binding: (1) sequence-specific binding patterns on genomes known as "motifs" and (2) interactions among TFs known as "co-binding effects". In this paper, we propose a novel deep architecture, the Prototype Matching Network (PMN) to mimic the TF binding mechanisms. Our PMN model automatically extracts prototypes for each TF through a novel prototype-matching loss. We use the notion of support set of prototypes and an LSTM to learn how TFs interact and bind to genomic sequences. On a TFBS dataset with 2.1 million genomic sequences, the PMN significantly outperforms baselines and validates our design choices empirically. Not only is the proposed architecture accurate, but also models the underlying biology.
  • TL;DR: We introduce a novel method to update the input conditioned on output label dependencies for multi-label classification
  • Keywords: multi-label classification, genomics, memory-augmented, deep learning