The Wuhan University scientists developed a fingerprint-embedding framework for drug-target binding affinity prediction, FingerDTA (Big Data Mining and Analytics, November 24, 2022). Traditional in vivo drug discovery uses living subjects to find new drugs and can be costly and time consuming. This new process is a virtual pre-screen that not only reduces costs but could improve the success rate in finding the right drug, the authors contend.

FingerDTA uses fingerprints, or descriptors, of drugs and targets. The targets are molecules related in some way to the diseases being pursued. Using general information from the drug or target fingerprint combined with a convolutional neural network model better predicts drug-target binding affinity.

Instead of the widely used MapReduce type programming model that often processes large amounts of data across hundreds or thousands of servers, the scientists used a more scalable method. This method developed by the Wuhan team reduces the data communication cost and enables approximate computing, which is less dependent on memory. It also allows for quickly sampling multiple random samples, directly executes serial algorithms on local random samples without data communications among the nodes, while facilitating big data exploration and cleansing.

The "Non-MapReduce" computing method simplifies big data computing and can save energy in cloud computing, the scientists added. Next, the team hopes to use the FingerDTA framework in big data platforms and input it in real applications.

Computerized approach may boost drug discovery
University of Texas at Dallas and Novartis Pharmaceuticals researchers have created a computer-based platform that could make the currently lengthy and...

Keen Therapeutics collaborates with Chemical.AI to boost drug R&D
Biotechnology startup Keen Therapeutics and artificial intelligence technology company Chemical.AI have announced a milestone in their collaboration....

Machine learning may accelerate solutions to address protein design challenges
Machine learning can be used to create protein molecules much more accurately and quickly than previously possible, with the potential for developing...

MIT team finds limitations of AlphaFold’s AI protein structures in drug discovery
Advances in the modeling of protein-ligand interactions are needed to realize the potential of the artificial intelligence (AI) AlphaFold database of...