PII-034 - INVESTIGATION OF DRUG COMBINATIONS IN CANCER THROUGH UNBIASED IDENTIFICATION OF RECURRING DRUG RESPONSE PATTERNS IN CELL LINES.

A. John, E. Ghose, H. Gao, L. Wang; Mayo Clinic, Rochester, MN, USA.

Background: Cancer is rarely treated with a monotherapy, instead using drug combinations. We investigated potential anti-cancer drug pairs through identification of correlated drug response patterns of independently acting agents from public datasets.

Methods: The normalized log (IC50) and AUC metrics of 250 individual drugs from the GDSCv2, CTRPv2, and PRISM datasets that were tested in 10 or more cell lines and appearing in two or more datasets were considered. Pairwise drug response metrics in all 3 datasets were calculated and corrected for multiple testing using the Benjamini-Hochberg (BH) procedure. Class-class associations were determined via distinct Anatomical Therapeutic Chemical (ATC)-codes showing enrichment with one-tailed Fisher’s exact tests performed for each ATC code in the target set against possible associations and corrected via BH with a significance cutoff of p-value < 0.05.

Results: 1: In all tissues-AUC-Spearman-BH condition, we found 2764 pairs to be significant at α = 0.05. The lowest calculated p-value was 1.39x10-120 was for vincristine and YK-4-279, an inhibitor of EWS-FLI1/RNA Helicase A, with a coefficient of 0.77. Over 15% of these pairs had adjusted p-values below 5x10-6.
2: Class enrichment on the same set showed 581 class pairs were enriched in the target set compared to the background set (n=1596 possible pairs) after BH correction. Notable enriched class combinations include: taxanes/Plk1 inhibitors (OR = 69.5, adjusted p = 5.6x10-11); Aurora Kinase inhibitors/histone modifying agents (OR=9.2, adjusted p = 6.13x10-11); pyrimidine analogues/CDK inhibitors (OR=3.43, adjusted p = 0.02).

Conclusion: Identification of drug-drug relatedness from public data represents an intuitive means of finding combination therapeutics with increased rigor and new underlying connections between drugs.

Narayan, R.S., Molenaar, P., Teng, J. et al. A cancer drug atlas enables synergistic targeting of independent drug vulnerabilities. Nat Commun 11, 2935 (2020).