# Reconciling multiple connectivity-based systems biology methods for drug repurposing

## TL;DR

This review unifies twelve drug repurposing methods by categorizing them into gene set-driven and network-driven approaches, clarifying their similarities and differences.

## Contribution

The paper systematically reconciles diverse connectivity-based drug repurposing methods into two structured categories with homogenized formulations.

## Key findings

- Drug repurposing methods can be grouped into gene set-driven and network-driven approaches based on their strategies.
- Gene set-driven methods use enrichment or projection to compare transcriptomic signatures and rank drug candidates.
- Network-driven methods use biological networks and algorithms to uncover system-level relationships between diseases and drugs.

## Abstract

In the last two decades, numerous in silico methods have been developed for drug repurposing, to accelerate and reduce the risks about early drug development. Particularly, following Connectivity Map, dozens of distinct data-driven methods have been implemented to find candidates from the comparison of differential transcriptomic signatures. Interestingly, there have been multiple proposals to integrate available knowledge using systems biology databases and adapted algorithms from the network biology research field. Despite their similarities, these methods have been formulated inconsistently over the years, even if some of them are fundamentally similar. The aim of this review is to reconcile these integrative methods, focusing on elucidating their common structures while underlining the specificities of their strategies. To achieve this, we classified those methods into two main categories, provided schematic workflow representations, and presented a homogenized formulation for each.

This review brings together twelve in silico methods for drug repurposing that integrate systems biology data, organizing them into two main categories: gene set-driven and network-driven approaches.Gene set-driven methods compare disease and drug transcriptomic signatures by mapping them onto predefined or data-driven gene sets, employing enrichment analyses or feature-space projections to quantify similarity or opposition.Across gene set-driven methods, there three main workflow steps are described: (i) identification of relevant gene sets or modules, (ii) selection of gene sets of interest based on the query signature, and (iii) calculation of modular or global connectivity scores to rank potential therapeutic candidates.Network-driven methods leverage biological interaction networks and network biology algorithms (such as propagation or shortest-path analysis) to reveal deeper, system-level relationships between disease and drug perturbations.

This review brings together twelve in silico methods for drug repurposing that integrate systems biology data, organizing them into two main categories: gene set-driven and network-driven approaches.

Gene set-driven methods compare disease and drug transcriptomic signatures by mapping them onto predefined or data-driven gene sets, employing enrichment analyses or feature-space projections to quantify similarity or opposition.

Across gene set-driven methods, there three main workflow steps are described: (i) identification of relevant gene sets or modules, (ii) selection of gene sets of interest based on the query signature, and (iii) calculation of modular or global connectivity scores to rank potential therapeutic candidates.

Network-driven methods leverage biological interaction networks and network biology algorithms (such as propagation or shortest-path analysis) to reveal deeper, system-level relationships between disease and drug perturbations.