Drug2ways: Reasoning over causal paths in biological networks for drug discovery

Abstract

Elucidating the causal mechanisms responsible for disease can reveal potential therapeutic targets for pharmacological intervention and, accordingly, guide drug repositioning and discovery. In essence, the topology of a network can reveal the impact a drug candidate may have on a given biological state, leading the way for enhanced disease characterization and the design of advanced therapies. Network-based approaches, in particular, are highly suited for these purposes as they hold the capacity to identify the molecular mechanisms underlying disease. Here, we present drug2ways, a novel methodology that leverages multimodal causal networks for predicting drug candidates. Drug2ways implements an efficient algorithm which reasons over causal paths in large-scale biological networks to propose drug candidates for a given disease. We validate our approach using clinical trial information and demonstrate how drug2ways can be used for multiple applications to identify: i) single-target drug candidates, ii) candidates with polypharmacological properties that can optimize multiple targets, and iii) candidates for combination therapy. Finally, we make drug2ways available to the scientific community as a Python package that enables conducting these applications on multiple standard network formats.

Causal reasoning over knowledge graphs leveraging drug-perturbed and disease-specific transcriptomic signatures for drug discovery

“…Furthermore, we make two assumptions in the design of the algorithm. Firstly, paths with cycles or a length greater than 7 edges between a given drug and disease are not considered, assuming that the effects exerted by paths beyond this length are less biologically relevant [ 23 ]. Secondly, we allow for at most one error between the transcriptomic data and a given path (see pseudocode of the algorithm in Fig 3 ) .…”

“…Knowledge graphs. We demonstrate our methodology using two established publicly available KGs that contain causal relations across drugs, proteins, and diseases: OpenBioLink KG [53] and a custom KG [23]. Both KGs are originally generated from a compedia of independent databases; thus, containing unique causal interactions depending on the source databases they include.…”

“…While traditionally, these algorithms were applied on small causal networks, they have recently begun to be applied on large-scale KGs, given the increasing availability of causal information, including proteins, drugs and phenotypes. For instance, a recent algorithm we published, drug2ways, reasons over all paths between a drug and a disease in a KG to predict the effect of the drug as the cumulative effect of all directed interactions between these two nodes [ 23 ]. Reasoning over all paths overcomes the limitation of earlier algorithms that exclusively account for shortest paths on protein-protein interaction networks, oversimplifying the effect exerted by one node on another, as all other paths between the two nodes are ignored [ 24 – 25 ].…”

Task-driven knowledge graph filtering improves prioritizing drugs for repurposing

“…In [6] the authors of the Hetionet graph database have utilized a custom path-based metric in combination of regularized linear regression for drug repurposing on Hetionet. Drug2Ways [20] leverages paths of causal relations on biomedical KGs for drug discovery while [21] uses neighbourhood information around compounds and diseases for drug repurposing. These methods use paths and neighbourhoods around compounds and diseases in the graph directly as a part of their model's training.…”

Causal reasoning over knowledge graphs leveraging drug-perturbed and disease-specific transcriptomic signatures for drug discovery

“…While traditionally, these algorithms were applied on small causal networks, they have recently begun to be applied on large-scale KGs, given the increasing availability of causal information, including proteins, drugs and phenotypes. For instance, a recent algorithm we published, drug2ways, reasons over all paths between a drug and a disease in a KG to predict the effect of the drug as the cumulative effect of all directed interactions between these two nodes (Rivas- Barragan et al, 2020). Reasoning over all paths overcomes the limitation of earlier algorithms that exclusively account for shortest paths on protein-protein interaction networks, oversimplifying the effect exerted by one node on another, as all other paths between the two nodes are ignored (Chindelevitch et al, 2012;Krämer et al, 2014).…”