AI Drug Shrinks KRAS Colon Tumors in Animals

An AI drug that shrinks KRAS colon tumors in animals has been developed by Oxford Drug Design, marking a significant preclinical success. The company announced the first in vivo evidence that its generative-AI engine can deliver a first-in-class small molecule for KRAS-driven colorectal cancer. This lead compound, conceived and refined by the company's dual GenAI/SynthAI platform, demonstrated clear tumor regression in two animal models, matching the efficacy of the benchmark mTOR inhibitor rapamycin while avoiding its dose-limiting toxicities. This breakthrough brings hope for many colon cancer patients with KRAS mutations who currently have few effective treatment options.

What breakthrough has Oxford Drug Design achieved in KRAS-driven colorectal cancer treatment?

Oxford Drug Design has developed a first-in-class small molecule using its generative-AI platform that effectively shrinks KRAS-driven colorectal tumors in preclinical studies. The novel compound demonstrates efficacy comparable to the benchmark drug rapamycin but without its associated toxicities, creating a potential new therapy for patients with KRAS-mutated cancers who do not respond to standard treatments.

Oxford Drug Design's AI platform created a novel small-molecule drug that caused significant tumor regression in animal models of KRAS-driven colorectal cancer. The compound demonstrated high efficacy, matching a benchmark drug, but with a much better safety profile, avoiding dose-limiting side effects observed with older therapies.

Data from a genetically-engineered mouse model mimicking human colorectal cancer were reported in PharmaTimes UK. Animals receiving the oral compound daily experienced a 60% increase in median survival. This result was statistically equivalent to the group treated with rapamycin but notably avoided the severe side effects, such as 15% weight loss and mucosal ulceration, that plagued the comparator arm. Further studies revealed a therapeutic window of at least ten-fold, a significant improvement over previous attempts to indirectly target mutant KRAS.

How the platform reached the clinic-ready candidate

Oxford Drug Design's discovery cycle blends two proprietary engines:

Module	Purpose	Output used in KRAS programme
GenAI	De-novo generation of 3-5 million synthetically feasible molecules per week	1,200 virtual hits predicted to bind leucyl-tRNA synthetase, an essential node in RAS-addicted translation control
SynthAI	Retrosynthetic scoring and active-learning feedback from micro-scale wet-lab data	42 compounds advanced to cell-based assays; ADME/Tox filters reduced to 12; final lead selected after in vivo PK/PD modelling predicted once-daily human dosing

The drug's target, leucyl-tRNA synthetase, was identified via a CRISPR screen as a critical vulnerability in KRAS-mutant cells. Instead of targeting the KRAS protein directly - a strategy with limited success - the team targeted a "network dependency." By inhibiting this enzyme, the drug disrupts the supply of charged tRNAs essential for the high rate of protein synthesis that cancer cells rely on. The AI platform designed novel molecules that bind to a unique pocket on the enzyme, showing high selectivity and leaving healthy cells unharmed.

Why existing benchmarks fall short in KRAS tumours

Conventional treatments like rapamycin often fail in KRAS-driven tumors because the mutations create escape routes that bypass the drug's effects. Oxford Drug Design's compound circumvents this resistance by acting further upstream. It starves the cancer cell of essential building blocks for protein synthesis, leading to the depletion of key oncoproteins (like MYC and cyclin D1) and ultimately triggering cell death. In advanced colorectal explant models (CMS-3 and CMS-4), the AI-designed agent reduced viable tumor mass by 72% over 14 days, while rapamycin achieved only an 8% reduction before its effects stalled.

"The relative strength of the lead candidate over rapamycin in mutated