PAIRWISE: Deep Learning-based Prediction of Effective Personalized Drug Combinations in Cancer.

Combination therapies offer promise for improving cancer treatment efficacy and preventing recurrence. Preclinical screening strategies can prioritize synergistic drug combinations. However, identifying optimal drug combinations tailored to specific cancer subtypes and individual patients is extremely challenging due to the vast number of possible combinations and tumor heterogeneity. To address this gap, we combined deep learning with transfer learning to incorporate prior scientific knowledge and predicted drug synergy based on tumor-specific transcriptome profiles. PAIRWISE explicitly modeled synergistic effects of drug combinations in cancer cell lines or individual tumor samples based on drug chemical structures, drug targets, and transcriptomes of inferred samples. Of note, PAIRWISE outperformed competing models with an AUROC (the area under the receiver operating characteristic curve) of 0.847 on held-out cancer cell lines. Moreover, when applied to an independent dataset of combinations with Bruton Tyrosine Kinase inhibitors (BTKi) in Diffuse Large B Cell Lymphoma (DLBCL) cell lines, PAIRWISE accurately predicted synergistic drug combinations with an AUROC of 0.720. To further confirm the robustness of PAIRWISE predictions, we selected 30 approved or investigational agents for DLBCL treatment and validated their synergy with BTKi across eight non-Hodgkin lymphoma cell lines. The synergistic predictions of PAIRWISE showed strong concordance with in vitro screening results. These findings highlight PAIRWISE's potential as a powerful in silico tool to prioritize candidate personalized drug combinations for further experimental validation.