The process of drug discovery and development is akin to the quest for the holy grail: long, arduous and full of failures. The one vital difference between the two is that drug discovery sometimes succeeds in the quest, but that itself is meeting a dead end.

Blockbuster medicines are facing the end of their patents and companies are complaining that the current drug development process is too expensive to sustain. The journey from the laboratory bench to patients’ bedside is a gruelling and laborious grind to the eventual fruition. It is estimated that the process of screening for viable molecules alone will take up to six years.

When it comes to drug discovery, time is money.

Enters artificial intelligence (AI), a recent hot topic in all disciplines of science. Researchers are keen to tap into the new-kid-on-the-block, with plans to reduce the timeline to a mere 12 months.

Thus, the US witnessed the birth of a consortium of US public-private partnership, called Accelerating Therapies for Opportunities in Medicine (ATOM) between pharmaceutical giant, GlaxoSmithKline (GSK), the Department of Energy (DOE) laboratories headed by Lawrence Livermore National Laboratories, and the National Cancer Institute (NCI).

Identifying drug targets quicker, smarter and cheaper

ATOM aims to leverage the power of high-performance computing to develop in silico models that can facilitate the process of drug discovery. Instead of testing one molecule at a time, supercomputers will screen millions of molecules within a short time frame to evaluate their related structural relationships.

"Drug development can fail for many reasons," says Arron Hingorani, a genetic epidemiologist, who developed an algorithm that screened genetic information against protein structure data and known interactions to identify potential drug targets. "However, a major reason is the failure to select the correct target for the disease of interest."

ATOM teammates (from left) Calad-Thomson, Brase, Paragas, and Baldoni stand among new supercomputing assets at Lawrence Livermore National Laboratories. Photo credit: Julie Russell/Lawrence Livermore National Laboratories
ATOM teammates (from left) Calad-Thomson, Brase, Paragas, and Baldoni stand among new supercomputing assets at Lawrence Livermore National Laboratories. Photo credit: Julie Russell/Lawrence Livermore National Laboratories

Traditionally, the process of drug discovery was conducted via an iterative chemical process to discover promising molecules before developing them into useful medicine.

Pharmaceutical companies employ hundreds, if not thousands, of in-house scientists to perform these complicated works. As evident by the failing drug pipelines, this traditional approach no longer works. Drug discovery has become, as what Daniel Cressey wrote in a Nature article, “a billion-dollar gamble that isn't delivering enough profitable products to market.”

However, the pharmaceutical industry did succeed in generating a huge volume of data from decades of drug discovery and development work. Most of this information was never fully utilised as scientists could only focus on a handful of promising drug candidates at one time. But with supercomputers and advanced modelling techniques, the time has come for platforms such as ATOM to flex its muscle.

“There are hundreds of thousands of failed molecules or molecules no longer of interest with information on structures, analogs, toxicity, and structure-function relationships,” says John Baldoni, senior vice president of platform technology and science at GSK. “Why would you not want to put them into the greater good? We feel there is an obligation that we have to patients in trials to share these failed compounds so we can develop better drugs faster.”

Revolutionary or self-deceiving bubble?

The word "AI" carries a futuristic aura which may have tricked many into believing its superior capability in solving almost every problem. While ATOM has established substantial anticipation over what the group can achieve, critics were quick to dismantle the buzz built up around the use of artificial intelligence in drug discovery.

In his Forbes article, Andrii Buvailo, head of e-commerce at Enamine Ltd, said attempts to use AI in the medical field is not new. In fact, the scientific community had tried tapping into AI in the 1970s and 1980s but failed miserably. It was not until the recent exponential growth in computing power (following Moore's law) that AI regained widespread interest to aid drug discovery.

Even so, there is yet a single drug in the market that is AI-inspired. The technology has not proven its worth in this field currently. But, given the relatively recent uptake of AI by major pharmaceutical companies, it is hoped that results will show over time.

The pharmaceutical industry succeeded in generating a huge volume of data from decades of drug discovery and development work.
The pharmaceutical industry succeeded in generating a huge volume of data from decades of drug discovery and development work.

Additionally, without the extensive “training” and being fed terabytes of relevant data, AI could never achieve, let alone master, the abilities that humans take for granted. This is also why it is challenging to deploy AI in drug discovery, as data is a precious asset to any pharmaceutical company – and these organizations will think twice before sharing data.

“Data for a pharmaceutical company today is its currency – it’s like pounds in the bank,” said Baldoni.

While GSK is willingly releasing data of more than a million compounds to ATOM, it is not guaranteed that other pharmaceutical companies will follow suit.

For Baldoni and his team, it is all about moving forward, “There is a recognition that the state-of-the-art computers in the national laboratories will soon be at pharma companies. We might as well start building the tools to use them. We’ll get ahead by a few years, and that is critical.”

Whether Baldoni and his team can live up to their promise of shortening the discovery process will be worthy of close inspection. MIMS

Read more:
The road to drug discovery
Artificial intelligence: a friend or foe – this next digital frontier is changing the medical industry
The artificial-intelligent doctor is still a dream – for now

Sources:
https://cen.acs.org/articles/95/i4/Pharma-partnership-applies-deep-learning.html?h=-1917987131
https://www.theguardian.com/breakthrough-science/2017/oct/23/artificial-intelligence-the-key-to-developing-effective-drugs-quickly
http://www.nature.com/news/2011/110302/full/471017a.html
https://www.forbes.com/sites/forbestechcouncil/2017/08/03/artificial-intelligence-in-drug-discovery-a-bubble-or-a-revolutionary-transformation/#7b3ccf914494
https://www.wired.com/2016/12/artificial-intelligence-artificial-intelligent/
https://www.wired.com/2017/03/supercomputers-stocking-next-generation-drug-pipelines/