Artificial intelligence is expected to change for good the way drugs are crafted.

Four years ago, the American biotechnology firm Insilico Medicine harnessed artificial intelligence to conceptualize a molecule designed to target a protein implicated in fibrosis. In a mere span of 46 days, this endeavor materialized, serving as a proof of concept due to the existence of multiple functional drugs for the targeted protein, which greatly enriched the AI's training dataset.

In June, Insilico Medicine inaugurated Phase 2 clinical trials involving humans for a drug that was conceived and engineered entirely by AI, with the U.S. Food and Drug Administration (FDA) granting Orphan Drug Designation to INS018_055 for the treatment of Idiopathic Pulmonary Fibrosis (IPF). 

In the second half of July, it identified 60 candidates of an inhalation solution for ISM001-055. The latter is the first anti-fibrotic small molecule inhibitor developed leveraging its proprietary AI drug discovery platform Pharma.AI for the treatment of IPF. 

Historically, drug discovery has been an arduous and protracted journey, marked by a plodding pace and significant financial burdens. Scientists traditionally embark on an expedition to decipher the causative factors behind a given ailment, often identifying a protein as the primary culprit.

More to read:
What can be more powerful than AI? A brain-machine symbiosis

Subsequently, they navigate through a labyrinthine assortment of tens of thousands of potential compounds capable of targeting the implicated protein.

From this vast pool, a few compounds exhibit promise warranting synthesis, and an even smaller subset advances to more in-depth scrutiny before potentially making their way to human clinical trials.

In the midst of screening over a million molecules, on average only one compound manages to proceed to late-stage clinical trials and eventually secures approval for utilization. This intricate odyssey, from the initial breakthrough to regulatory endorsement, demands a staggering investment of 12 to 15 years and an approximate budget of $1 billion.

With artificial intelligence, which is at the core of research and craftsmanship, scientists can shorten the time, cut down the costs, and enhance efficiency, thus helping cure 300 million people worldwide who suffer from rare diseases, the company said on its blog. 

Employing AI

Insilico Medicine has deviated from those customary trajectories. Its AI-forged drug contender for idiopathic pulmonary fibrosis, a chronic ailment characterized by lung scarring and breathing difficulties, emerged within a third of the usual timeframe and at a mere fraction of the customary cost, thanks to the company's innovative technological arsenal.

The approach hinges on the synergy of two distinct AI methodologies.

The first is a generative adversarial network (GAN), wherein two neural networks engage in a duel of sorts. One network generates output while the other evaluates the veracity of said output. Through this interplay, these networks generate novel entities, ranging from textual constructs to visual depictions, and in the case at hand, chemical structures of minuscule molecules.

More to read:
Studies on longevity describe how five species found a pathway to reverse aging

Furthermore, Insilico's platform leverages reinforcement learning, a facet of machine learning that endows a system with the capacity to learn through iterative experimentation, guided by feedback garnered from its own actions. This form of learning has been pivotal in the advancement of AI systems specialized in gaming.

INS018_055 is a result of this innovative approach. Functioning as an anti-fibrotic small molecule inhibitor, this drug serves to impede the progression of tissue thickening and scarring, particularly within the lungs of patients. Presently, Insilico will test the drug in 60 individuals afflicted by IPF in both the United States and China, for a duration of 12 weeks. 

Around 5 million people suffer from IPF. People diagnosed with this ailment typically face a lifespan of merely three to four additional years.

According to an analysis compiled by Morgan Stanley, the integration of AI tools holds the potential to engender the creation of 50 novel drugs, collectively valued at approximately $50 billion within the forthcoming decade.