The first project, PaccMann, predicts anticancer compound sensitivity using "multi-modal attention-based neural networks". Big Blue is working on the PaccMann algorithm to analyse chemical compounds and predict which are the most likely to fight cancer strains.
The ML algorithm exploits data on chemical compound gene expression and molecular structures. Big Blue says that by identifying potential anti-cancer compounds earlier, this can cut the costs associated with drug development.
The second project has the catchy title "Interaction Network infErence from vectoR representATions of words" [INtERAcT]. This tool automatically extracts data from valuable scientific cancer related papers. INtERAcT makes the academic side of research easier by automatically extracting information from these papers. The tool extracts protein-protein interaction data. This is an area of study which could mess up the biological processes in diseases – including cancer.
The third project is "pathway-induced multiple kernel learning", or PIMKL. This algorithm uses datasets describing what we know when it comes to molecular interactions to predict cancer progression and relapses in patients. PIMKL uses multiple kernel learning to identify molecular pathways crucial for categorising patients, giving healthcare professionals an opportunity to individualize and tailor treatment plans.