Abstract
Incorrect drug target identification is a major obstacle in drug discovery. Only 15% of drugs advance from Phase II to approval, with ineffective targets accounting for over 50% of these failures1–3. Advances in data fusion and computational modeling have independently progressed towards addressing this issue. Here, we capitalize on both these approaches with Rosalind, a comprehensive gene prioritization method that combines heterogeneous knowledge graph construction with relational inference via tensor factorization to accurately predict disease-gene links. Rosalind demonstrates an increase in performance of 18%-50% over five comparable state-of-the-art algorithms. On historical data, Rosalind prospectively identifies 1 in 4 therapeutic relationships eventually proven true. Beyond efficacy, Rosalind is able to accurately predict clinical trial successes (75% recall at rank 200) and distinguish likely failures (74% recall at rank 200). Lastly, Rosalind predictions were experimentally tested in a patient-derived in-vitro assay for Rheumatoid arthritis (RA), which yielded 5 promising genes, one of which is unexplored in RA.
Highlights
Incorrect drug target identification is a major obstacle in drug discovery
In order to condition the model to learn to predict those drug targets most likely to be therapeutically linked to a disease, Rosalind used a subgraph consisting of Disease-GeneProtein links with relation ‘Therapeutic Relationship’ as a benchmark
The performance of the scoring function used by Rosalind, ComplEx, was compared with three comparable scoring functions used in matrix factorization methods: DistMult, canonical polyadic factorization (CP), and holographic embeddings (HolE)
Summary
Incorrect drug target identification is a major obstacle in drug discovery. Only 15% of drugs advance from Phase II to approval, with ineffective targets accounting for over 50% of these failures[1,2,3]. We capitalize on both these approaches with Rosalind, a comprehensive gene prioritization method that combines heterogeneous knowledge graph construction with relational inference via tensor factorization to accurately predict disease-gene links. Over the past 200 years, only about 1,500 drugs cleared clinical trial and reached approval, leaving the majority of nearly 9,000 diseases without the possibility of treatment options[5] These failure rates incur a huge financial and societal cost, and highlight a need for improved approaches to selecting more effective drug targets at the early development stage, a process known as gene prioritization[6]. We introduce a novel method for gene prioritization, Rosalind, that combines relational inference via tensor factorization with graph-based data integration to predict disease genes. Rosalind is able to make prospective predictions using time-sliced d ata[15] as well as predict those genes that have a high probability of efficacy in a clinical trial[16]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
