x

AI based drug-response prediction using multi-omics

    BioICAWTech > AI based drug-response prediction using multi-omics

AI based drug-response prediction using multi-omics

Multi-omics data integration has become a potent strategy for improving the precision of drug response prediction models. We have recently developed AI based prediction method that is able to accurately predict drug responses for any compound or approved drug across > 1000 cancer cell lines. Our AI method was trained based on optimized multi-omics descriptors (i.e. gene-expressions, mutation, methylation and copy number variations) and structure-based descriptors (i.e. several structural fingerprints and physiochemical properties).

Prediction method was tested on several independent datasets from public platforms such as NCI-60 and CCLE and shown significant performance. We also experimentally validated several predicted endpoints and currently in the process to further improve the drug-response prediction algorithm.

Meanwhile, we have recently developed a highly customized web-portal that visualizes bi-partite networks by harmonizing drug sensitivity data from major drug response platforms such as: CCLE, GDSC, NCI-60, gCSI, FIMM and NCATs. It is available at: https://shop.icawtech.com/drugtargetnetwork/index.php.There are nearly 2000 cell lines and 50K compounds/drugs in our portal. The web portal developed by our company is unique in terms of following points:

  • It harmonizes drug-sensitivity data from all platforms at place and provide customized filtration based on tissue types, disease sub types, compound’s clinical status, drug sensitivity range, drug-target information and many more.
  • Portal also provided AI based prediction method that can predict drug-sensitivity for any compound across 1000 cancer cell lines. Users just need to provide compound’s SMILE.
  • Provides novel experimentally generated data for approved drugs across number of cancer cell lines (ongoing)
  • High customized bi-partite network visualization that can be freely exported and used by users or scientists in their manuscripts/presentations.

Following Figure shows exported bi-partite visualization network generated by our portal for approved and phase II drugs across FIMM and gCSI data portals for Breast, lung, pancreas, prostate, bone and lymphoid based cell lines.