Machine Learning in Pharmaceuticals
- (The Frick Chemistry Laboratory, Princeton University - Kimberly Chen)
- Overview
Machine Learning (ML), one of the most prominent approaches in artificial intelligence, is the future of the pharmaceutical industry. The Human Genome Project and subsequent thousands of discoveries at the DNA, RNA, and protein levels were made possible by ML's ability to detect patterns in large and often messy datasets. ML has the potential to accelerate the clinical drug discovery and development process by applying sophisticated algorithms to the analysis and mining of disparate data sources to predict molecular behavior and suitability as drug targets or therapeutic entities.
ML plays a vital role in enhancing the work of drug development researchers so that informed preliminary analysis of large amounts of scientific data can be performed to form necessary new knowledge. What was once a purely hypothesis-driven approach to human-asked questions is shifting towards scientists starting with a result and using machine learning to help discover important relationships in the data to that result.
ML will also help the industry select patients for clinical trials and enable companies to detect any issues with compounds earlier when it comes to efficacy and safety. Therefore, the industry can gain many benefits by adopting an ML approach. It can be used to build a robust, sustainable pipeline of new drugs with good results.
- The Problems of the Current Drug Discovery Process
The current drug discovery process - too long and very expensive. It can take up to 15 years to translate a drug discovery idea from an initial start into a market-ready product. Industry is currently said to be spending well over $1 billion per drug. That's partly because all drugs that don't work out have to be paid for.
The drug discovery process is changing as our understanding of biology deepens due to the availability of new data and algorithms from which to learn. Machine learning presents a real opportunity for the pharmaceutical industry to conduct R&D differently, allowing it to operate more efficiently and significantly improve success rates in the early stages of drug development.
- The Machine Learning in the Drug Discovery Process
Recent innovations in machine learning techniques can greatly aid researchers in the drug discovery process and advancing the pipeline. The average biomedical researcher is dealing with vast amounts of new information every day. An estimated 10,000 new publications are uploaded every day in the bioscience industry - from all over the world and from a wide variety of biomedical databases and journals.
Therefore, it is impossible for researchers to know, let alone process, all the scientific knowledge relevant to their field of study. What's more, without the ability to correlate, assimilate, and connect all of this data, it would be impossible to create new usable knowledge -- which could be used to develop new drug hypotheses.
[More to come ...]
