GSK puts faith in AI to make more successful drugs more quickly | Business
However, the UK’s largest drugmaker by revenue is struggling to hire enough AI researchers and engineers from areas such as Silicon Valley and is looking to former employees in academia, the US Navy and the music industry to fill positions in the new team. They will be spread across London, Heidelberg, San Francisco, Philadelphia and Boston.
The AI unit will be headquartered in San Francisco, with one GSK executive admitting competition for AI professionals is fierce. “In AI, we are scouring the planet for the best people. These folks are very rare to find. Competition is high and there aren’t a large number of them,” said Tony Wood, GSK’s senior vice-president of medicinal science and technology.
Among those to be hired will be researchers qualified at PhD or master’s level who will be working on AI drug development projects in London under a new fellows programme. AI will be used to help find treatments for conditions such as cancer and autoimmune diseases including rheumatoid arthritis.
AI refers to the use of trained algorithms to carry out tasks, which computers have learned to do through repetitive processes rather than instruction from humans. In recent years, the increasing use of AI in the pharmaceutical industry has begun to transform the way in which scientists develop new drugs. Drug development traditionally takes years, but AI can speed the process up and increase the success rate of new medicines from 10% currently.
GSK believes choosing targets – proteins or other molecules – based on genetic evidence means they are twice as likely to succeed. This has led the company to invest heavily in human genome research, which involves analysing the workings and interactions of genes, and cell and gene therapy, which aims to fix genetic defects and re-engineer patients’ cells to fight disease.
Wood said: “Functional genomics looks at hundreds of millions of data points from genetic databases, which was until recently out of reach of data science.”
But the data can now be analysed thanks to increased computer power and the use of deep learning, a type of machine learning in AI that enables better analysis through integration of different types of data, such as cell imaging and genomic data.
The main base of GSK’s AI team will be next to a new laboratory focused on functional genomics in San Francisco that the drugmaker is building in partnership with the University of California.
GSK developed one of the first gene therapies, Strimvelis, for a rare immune deficiency known as “bubble baby” syndrome.
GSK has since sold Strimvelis as part of its rare disease portfolio and is focusing on cancer treatments, with 16 in development. The most advanced is NY-ESO-1, a T-cell therapy, for solid tumours. Larger clinical trials will start in 2020 at a hospital near London for patients with synovial sarcoma, a rare type of cancer that affects soft tissues. Other studies are evaluating its use in lung and other cancers.
In T-cell therapy, white blood cells called T-cells are taken from a patient. They are reprogrammed so they attach on to cancer cells and attack them when they are reinjected into the patient.
So far, two cell therapies have been approved for blood-borne cancers worldwide – Kymriah for leukaemia by Switzerland’s Novartis, and Yescarta for lymphoma by Gilead of the US – but there are none for common solid tumours.