AI won’t be the much-prophesised saving grace that rescues the pharmaceutical industry from existential decline argues Drew Smith, in a recent Medium blog. Despite spewing out exponentially more drug candidates than traditional discovery approaches, this isn’t where the big costs lie.
Decades of technological and scientific advances since the 1950s pharma glory days when your average billion dollars spent in R&D returned twice as many approved drugs, should have brought with it greater efficiency. And, whilst molecule discovery is much more powerful today, it has only been coupled with rocketing development costs. In fact, they increase an average of 9% year on year.
The problem is that we hit all the “gold nuggets”, the easy targets years ago. So far, AI hasn’t got us much closer to the remaining specks of druggable targets. With most new approvals in the arena of orphan indications, affecting less than 200,00 patients, the outlook for future blockbusters, the next Valium or Viagra, seems bleak.
Despite the litany of AI tech companies offering discovery platforms, we’re yet to see solutions that revolutionise target discovery. And this is where the true problem lies: the largest R&D costs today stem from an intrinsic lack of druggable, high affinity targets.
Instead, where AI might yet prove it’s worth is in tackling another long-standing drug development obstacle: intrinsically disordered proteins (IDPs). Regarded as the shape shifters of biology this group of proteins have been implicated in driving a plethora of conditions, from cancer to neurodegenerative disease, but remained impossible to target. Their continually changing configurations and elusivity in targeting have earned them the moniker “the undruggables”.
The computational approaches developed to target these proteins, including the tumour suppressor P53, are divided into four categories: physiochemical-based, machine learning-based, template-based, and meta method-based, and have been recently reviewed here.
Setting potential IDP-targeting drugs aside, Smith concludes that the much-anticipated pharma revolution ushered in by the dawn of artificial intelligence won’t be a revolution at all. It’s success he wagers, will be limited to the realm of niche. Not blockbusters. A revolution for the individual, not the masses.
