A Real Example, and What It Shows | AI Learning Guide

The rentosertib example matters because it is the clearest case of the whole pipeline running end to end with AI as a central tool — target, molecule, testing, and eventually human trials, in about two and a half years rather than the four to six that is typical. But be careful about what that proves. One candidate reaching trials shows the approach can work; it does not show that it is faster or better in general. And for cancer specifically, the honest position is that AI-designed drugs are in testing with some promising early results, but there is not yet a widely accepted example of an approved cancer drug that came this way and changed how patients are treated.

0:00 / 0:00

Auto-play

Rentosertib: an AI-driven candidate that reached trials

Insilico Medicine used AI to propose a target, generate molecules against it, and select a candidate, which then entered human trials in roughly two and a half years. The disease was idiopathic pulmonary fibrosis rather than cancer, but the process followed the same stages this course has described. The milestone is that the full pipeline ran end to end with AI as a central tool and produced a molecule that reached patients in a trial.

A single candidate reaching trials is evidence that the approach can work, not proof that it is faster or better in general. The honest reading is that the method has cleared an important hurdle while the broader claim — that AI reliably shortens cancer drug discovery — remains unproven.

For cancer specifically, AI-discovered or AI-optimized drugs are in clinical testing and some have shown promising early results, but no widely accepted example of an approved, practice-changing AI-designed cancer drug exists yet.