Machine learning models for drug discovery
IBM researchers Ping Zhang (left) and Jianying Hu (right) have been granted U.S. Patent 9,536,194 for an invention that could accelerate discovery of more effective and safer drugs by using machine learning models to predict therapeutic indications and side effects from various drug information sources. Credit: IBM
IBM today announced that its scientists have been granted a patent on machine learning models to predict therapeutic indications and side effects from various drug information sources. IBM Research has implemented a cognitive association engine to identify significant linkages between predicted therapeutic indications and side effects, and a visual analytics system to support the interactive exploration of these associations.
This approach could help researchers in pharmaceutical companies to generate hypotheses for drug discovery. For instance, strongly correlated disease-side-effect pairs identified by the patented invention could be beneficial for drug discovery in many ways. One could use the side-effect information to repurpose existing treatments (e.g. drugs causing postural hypotension could be potential candidates for treating hypertension). If a new drug is being designed for a disease that is strongly correlated with severe side effects, then special attention could be paid to controlling the formulation and dosing of the drug in the clinical trials to prevent serious safety issues.
IBM was granted U.S. Patent 9,536,194: Method and system for exploring the associations between drug side-effects and therapeutic indications for this invention.
Lack of efficacy and adverse side effects are two of the primary reasons a drug fails clinical trials, each accounting for around 30 percent of failures. Computational models and machine learning methods that can derive useful insights from large amounts of data on drugs and diseases from various sources hold great promise for reducing these attrition rates and improving the drug discovery process.