# RWDisEnh+: Enhancing disease-enhancer association prediction through multiplex-heterogeneous networks

## TL;DR

RWDisEnh+ is a new computational method that improves predictions of how enhancers are linked to diseases by combining gene-based and sequence-based data in a network framework.

## Contribution

RWDisEnh+ introduces a sequence-based enhancer similarity network into a multiplex-heterogeneous framework to enhance disease–enhancer association predictions.

## Key findings

- RWDisEnh+ achieves an average AUC of 0.874, outperforming the previous version RWDisEnh.
- The method identifies more evidence-supported disease–enhancer associations, including 10 enhancers linked to seven diseases.
- Predicted associations are enriched in immune, inflammatory, and metabolic pathways.

## Abstract

Enhancers are critical regulatory DNA elements that, when dysregulated, can contribute to disease pathogenesis by altering gene expression. Although millions of enhancers have been identified through large-scale genomic projects, their associations with human diseases remain largely uncharacterized, emphasizing the need for robust computational approaches. In our previous work, we developed RWDisEnh, a network-based method that integrates a shared gene–based enhancer network with a disease similarity network within a heterogeneous framework to predict disease–enhancer associations. In this study, we present RWDisEnh+ , an enhanced version of RWDisEnh that incorporates a sequence-based enhancer similarity network into a multiplex-heterogeneous network to improve prediction performance. Using an extended random walk with restart (RWR) algorithm, RWDisEnh+ allows information to propagate across disease and enhancer layers, leveraging both gene-based and sequence-based similarity features to rank candidate enhancers for each disease. Comprehensive evaluation using 3-fold cross-validation demonstrated that RWDisEnh+ achieves an average AUC of 0.874, outperforming RWDisEnh’s AUC of 0.819. Moreover, RWDisEnh+ identifies a larger number of evidence-supported disease–enhancer associations across top-k rankings, including 10 enhancers linked to seven diseases such as asthma, rheumatoid arthritis, and type 2 diabetes. GWAS validation and pathway enrichment analyses further reveal that these predicted associations are enriched in immune, inflammatory, and metabolic pathways, highlighting their biological relevance. Overall, RWDisEnh+ provides a stable and effective framework for predicting novel disease–enhancer associations, offering new insights into enhancer-mediated gene regulation and the genetic architecture of complex diseases.