Multi-objective screening of non-small cell lung cancer drug candidates

Despite improvements in diagnosis and chemotherapy, non-small cell lung cancer (NSCLC) remains one of the most common cancer and has the largest proportion of all cancer death rates today. Computational approaches have been widely applied for early detection of novel treatment for NSCLC. Herein we developed a multi-objective approach or the screening of chemical compounds simultaneously active against three NSCLC cell lines: A549, NCI-H1299 and NCI-H1975. The first step consisted of developing ensemble models based on cytotoxicity data against three NSCLC cell lines curated from ChEMBL database. A desirable-based algorithm was then applied to incorporate these models into a multi-objective optimization system that can be used for virtual screening protocol. This system showed suitable screening performance with the Boltzmann-Enhanced Discrimination of ROC BEDROC = 0.62, the Enrichment Factor (EF)_1% = 30 and the Area Under the Accumulation Curve (AUAC) = 0.69