Inductive Bio Wins Third Consecutive AI Drug Discovery Benchmark, Topping 350+ Competitors in PXR Blind Challenge
核心洞察
Inductive Bio (搜索) placed first in OpenADMET's PXR Blind Challenge, outperforming more than 350 researchers from large pharma, biotech, and academic organizations.
The company's Beacon AI model has now won three consecutive independent blind challenges, a record unmatched by any other AI drug discovery company.
The PXR challenge focused on predicting pregnane X receptor activation, a common liability that often goes undetected until mid- to late-stage lead optimization.
Inductive Bio (搜索), an AI drug discovery company developing virtual chemistry labs, has placed first in OpenADMET's PXR Blind Challenge, beating more than 350 researchers across large pharmaceutical, biotechnology, academic, and AI organizations. The victory, announced July 14, 2026, marks the company's third consecutive win in the industry's leading AI small molecule competitions, establishing a track record unmatched by any other AI drug discovery firm.
The PXR Blind Challenge tasked participants with predicting activation of the pregnane X receptor (PXR) (搜索), a protein that detects foreign compounds and triggers the body to metabolize and eliminate them. PXR induction represents a common liability that frequently goes undetected until mid- to late-stage lead optimization, often years after discovery efforts have begun. The blind challenge format, in which participants predict properties of previously unseen compounds, provides a rare opportunity to benchmark ADMET (Absorption, Distribution, Metabolism, Excretion, and Toxicity) modeling under real-world conditions.
"PXR induction is one of those problems that kills programs late, after teams have already devoted significant resources to a compound series," said Ben Birnbaum, Inductive's Co-Founder and CTO. "Predicting it means picking up on subtle signals in chemical structure. This first-place finish tells us that our Beacon models are catching those signals, showing how far these models can go on the toughest drug discovery problems."
Unmatched Performance Across Three Blind Challenges
Across the three blind challenges, over 750 competitors made nearly 10,000 submissions spanning a total of 17 critical assay endpoints for drug discovery. Inductive placed first in all three challenges, outperforming entrants from companies many times its size, including Merck (搜索) (with NVIDIA (搜索)), Novo Nordisk, EMD Serono, and others. The company's prior wins include top honors in the ASAP-Polaris-OpenADMET Antiviral Challenge and the ExpansionRx-OpenADMET blind challenge.
The Beacon models are trained on one of the industry's largest and most diverse datasets, drawn from Inductive's pre-competitive data consortium, and fine-tuned to each partner's chemical space. The company's virtual lab currently powers the discovery programs of dozens of biopharma partners and has delivered multiple development candidates in significantly shorter timelines than the industry standard.
Broader Drug Safety Mission and ARPA-H Award
Inductive's work on PXR is part of a broader drug safety focus. The company was selected to lead an up to $21 million ARPA-H CATALYST award, alongside Amgen and academic partners, to build AI models of drug-induced liver injury and cardiotoxicity — two of the largest drivers of clinical drug safety failures.
"We built Inductive to give every drug hunter a virtual chemistry lab, so teams can predict how a molecule will behave before spending months testing it in a wet lab and focus their effort on only the highest-quality ideas," said Josh Haimson, Inductive's co-founder and CEO. "Three consecutive wins in independent blind challenges against teams 1000x our size tell us that the Beacon models powering that lab are the state of the art for AI drug discovery tasks like ADMET prediction."
Inductive's virtual labs are designed to scale proven scientific best practices across medicinal chemistry, computational chemistry, DMPK, and safety, enabling teams to make higher-quality decisions consistently throughout discovery. AI chemistry assistants work alongside predictive ADMET and PK models to help scientists evaluate more hypotheses in silico and surface key risks earlier, with the most promising molecules advancing from the virtual lab to the wet lab in a tight feedback loop.