Integrated Tools for Enhancing Network-based Disease Module Identification, Drug Prioritization and Repurposing

The domain of network-based drug repurposing research encounters considerable challenges in disease module identification and evaluation, further complicated by the absence of a gold standard for assessment and the requirement for user-friendly tools. The identification process involves extracting interactome genes with contextual similarities to a predefined condition-specific gene set, a process highly dependent on the use case and algorithm selected. The current tools often exhibit minimal overlap due to varied foundational assumptions, leading researchers to select tools without comprehensive suitability assessment, resulting in a preference for tools yielding expected results but potentially failing to uncover new, actionable drug repurposing candidates.

To address the lack of robustness and the bias in tool selection, we introduce CAMI (Consensus Active Module Identification), a consensus approach suite combining existing algorithms like DOMINO [1], DIAMOnD [2], and ROBUST [3] while permitting further extensions. The ensemble classifier in CAMI is adaptable and, by default, rivals the performance of the best base tool across multiple scenarios. We assessed its efficacy using the seed rediscovery rate, a metric computed by sequentially removing a fraction of input genes (seeds) and determining the ratio of discovered, excluded genes to the module size.

Indirect performance metrics are essential for module evaluation. Hence, we developed DIGEST, a Python-based tool for assessing disease module coherence that streamlines the in silico validation process via fully automated validation pipelines [4]. These encompass disease and gene ID mapping, enrichment analysis, shared gene and variant comparisons, and background distribution estimation, enabling users to evaluate the statistical significance of candidate modules or extracted connected components serving as candidate mechanisms concerning functional and genetic coherence and to effortlessly compute empirical P-values.

In summary, CAMI and DIGEST, accessible as Python packages, provide an all-encompassing solution for network-based drug repurposing research. They deliver a robust approach for disease module identification and validation, diminishing tool selection bias and facilitating indirect performance assessment. While DIGEST is available as a web tool ( https://digest-validation.net/ ), a CAMI web tool is currently under development.