Motif Diversity in Human Liver ChIP-seq Data Using MAP-Elites

TL;DR

This paper reframes DNA motif discovery as a quality-diversity problem and applies MAP-Elites to evolve fixed-length PWMs under a likelihood-based objective, yielding an archive of diverse, high-quality motifs rather than a single best motif. The fitness is defined by a log-odds discriminative score, with $f(M) = \frac{1}{k} \sum_{s \in Top_k(S; g)} g(M,s)$ and $g(M,s) = \frac{1}{L} \max_i \max_{d \in \{+,-\}} \ell(M, s^{(d)}_{i:i+L-1})$, where $\ell(M,w) = \sum_{j=1}^L \log \frac{P_M(w_j|j)}{P_{bg}(w_j)}$. Through three descriptor pairings (ME.SP, ME.CO, ME.RB), MAP-Elites reveals motif variants that balance specificity, compositional structure, and robustness, achieving fitness comparable to MEME’s top motifs while exposing a structured diversity hidden by single-solution methods. The results suggest that quality-diversity illumination provides a practical and interpretable framework for motif analysis in regulatory genomics, enabling richer biological interpretation across contexts. Future work should extend to more factors, longer motif lengths, and additional QD algorithms to scale this approach.

Abstract

Motif discovery is a core problem in computational biology, traditionally formulated as a likelihood optimization task that returns a single dominant motif from a DNA sequence dataset. However, regulatory sequence data admit multiple plausible motif explanations, reflecting underlying biological heterogeneity. In this work, we frame motif discovery as a quality-diversity problem and apply the MAP-Elites algorithm to evolve position weight matrix motifs under a likelihood-based fitness objective while explicitly preserving diversity across biologically meaningful dimensions. We evaluate MAP-Elites using three complementary behavioral characterizations that capture trade-offs between motif specificity, compositional structure, coverage, and robustness. Experiments on human CTCF liver ChIP-seq data aligned to the human reference genome compare MAP-Elites against a standard motif discovery tool, MEME, under matched evaluation criteria across stratified dataset subsets. Results show that MAP-Elites recovers multiple high-quality motif variants with fitness comparable to MEME's strongest solutions while revealing structured diversity obscured by single-solution approaches.

Figures (2)

• Figure 1: MAP-Elites archive structure for a representative CTCF ChIP-seq subset. Heatmaps show elite motif fitness under three behavioral characterizations; white cells indicate unpopulated regions of the descriptor space.
• Figure 2: Representative motifs discovered on the same subset. Best-performing motifs from MAP-Elites archives (a-c) and MEME (d), with fitness values $0.790$, $0.950$, $0.722$, and $1.069$.