Machine Learning in Pharmaceutical Research and Development

Lets look at the Pharmaceutical Industry for a minute... 

Its a complex and competitive world out there for a multitude of reasons: There are time limits on patented treatments, customers have more bargaining power, and we are moving on to tackle more complex diseases.

The more complex the disease, the more complex the research and development (R&D) phase for treatments. The R&D pipeline for treatments of rare, chronic, degenerative diseases is lengthy and costly, with many false starts and do-overs involved. How does it work? A simplified version is that researchers test the interaction of different molecules to gauge is a combination would be effective in killing the molecules of the disease in question. The process is traditionally done by human researchers using vast amounts of data about the properties of molecules to see the interaction between multiple molecules, and then their interaction with the disease.  As you can guess, its an extremely time consuming and costly process..

But with Machine Learning (more specifically deep learning and neural networking) those interactions are now able to occur in silico, meaning done via computer simulation.  Before deep learning, the way computers tackled the property interactions was done methodically and sequentially but as the computers learn the simulations occur in a much more fluid way.  Multiple properties are tested at once and that new, updated, information reduces the need to test molecules with the same properties.  The computer has already learned what that combination does!

This speedy new process is revolutionizing the way we approach drug discovery.  It dramatically shortens the R&D process and is speculated to save 30% of typical costs. Hopefully this is good news, hopefully it leads to better and more accessible treatments for people who need it.

Want to learn more? We checked out these articles to learn more about this topic, explore!

Artificial intelligence yields new antibiotic
by Anne Trafton, MIT News Office
Feb 20, 2020

AI tool screens 107 million molecules, discovers potent new antibiotics
by Jamie Durrani, Chemistry World
Feb 21, 2020