# Drug Discovery

27 related articles

Subscribe to updates

德睿智药（MindRank） Nov 13, 2026
MindRank and Borui Cuanghe team up on cyclic-peptide RDCs: InteraFold-PEP predicts binding modes before any candidate exists
MindRank and Borui Cuanghe have announced a strategic collaboration to advance AI-driven radionuclide drug conjugates (RDCs) that use cyclic peptides as targeting ligands, aimed at unmet needs in precision oncology. MindRank contributes its Intera life-science foundation model system, with InteraFold-PEP predicting how cyclic peptides bind target proteins to guide design, screening and optimization; Borui Cuanghe contributes radiopharmaceutical development, radiolabeling and clinical translation, and the two sides iterate AI design against experimental validation. The target ligands must combine high affinity, high selectivity and good developability, and the work is at the collaboration-launch stage.
MindRank and Borui Cuanghe have announced a strategic collaboration to advance AI-driven radionuclide drug conjugates (RDCs) that use cyclic peptides as targeting ligands, aimed at unmet needs in precision oncology. MindRank contributes its Intera life-science foundation model system, with InteraFold-PEP predicting how cyclic peptides bind target proteins to guide design, screening and optimization; Borui Cuanghe contributes radiopharmaceutical development, radiolabeling and clinical translation, and the two sides iterate AI design against experimental validation. The target ligands must combine high affinity, high selectivity and good developability, and the work is at the collaboration-launch stage.
德睿智药（MindRank）
Strategic Collaboration Intera life science foundation model Drug Discovery
德睿智药（MindRank） Nov 13, 2026
MindRank plugs its Intera life-science foundation model into Tencent's WorkBuddy-based AI workbench, putting target discovery through clinical research in one workflow
MindRank and Tencent Health announced a technical collaboration at the 2026 Tencent Global Digital Ecosystem Summit to bring MindRank's in-house Intera life-science foundation model into Tencent's one-stop life-science AI workbench, built on WorkBuddy. As one of the workbench's first ecosystem partners, MindRank plans to fold Intera into drug R&D workflows spanning target discovery, molecular design, virtual screening, automated synthesis, biological testing and clinical research, adding models versed in multiple drug modalities. The two sides will also explore structured return of experimental and clinical data to build a clinical data-in-the-loop cycle that keeps iterating the models.
MindRank and Tencent Health announced a technical collaboration at the 2026 Tencent Global Digital Ecosystem Summit to bring MindRank's in-house Intera life-science foundation model into Tencent's one-stop life-science AI workbench, built on WorkBuddy. As one of the workbench's first ecosystem partners, MindRank plans to fold Intera into drug R&D workflows spanning target discovery, molecular design, virtual screening, automated synthesis, biological testing and clinical research, adding models versed in multiple drug modalities. The two sides will also explore structured return of experimental and clinical data to build a clinical data-in-the-loop cycle that keeps iterating the models.
德睿智药（MindRank）
Intera life science foundation model AI drug discovery platform integration Drug Discovery
arXiv Nov 11, 2026
BindMol splits the binding pocket into hotspots: higher novel hit rates than GEAM on five targets and more frequent recovery of type-I/II binding modes
Generative models for target-based drug design have long compressed protein–ligand interactions into a single docking score, hiding functional differences between binding modes on the same target. BindMol clusters the binding pocket into hotspots from reference complexes, assigns each hotspot its own reward dimension, and optimizes the conflicting objectives with preference-conditioned multi-objective reinforcement learning. On parp1, fa7, 5ht1b, braf and jak2 its novel hit ratios are 42.617, 24.580, 42.149, 30.767 and 44.864, above GEAM at 40.635, 17.903, 36.667, 27.887 and 42.025, and it recovers the crystallographic type-I and type-II binding modes more often on braf and jak2.
Generative models for target-based drug design have long compressed protein–ligand interactions into a single docking score, hiding functional differences between binding modes on the same target. BindMol clusters the binding pocket into hotspots from reference complexes, assigns each hotspot its own reward dimension, and optimizes the conflicting objectives with preference-conditioned multi-objective reinforcement learning. On parp1, fa7, 5ht1b, braf and jak2 its novel hit ratios are 42.617, 24.580, 42.149, 30.767 and 44.864, above GEAM at 40.635, 17.903, 36.667, 27.887 and 42.025, and it recovers the crystallographic type-I and type-II binding modes more often on braf and jak2.
Dingshuo Chen（First author）
BindMol Envelope SAC Multi-objective reinforcement learning Energy-based fragment scoring Drug Discovery
arXiv Nov 5, 2026
PharmAgents2 hands drug discovery to agents: it proposes a new LpxH allosteric site and the compounds pass wet-lab checks
PharmAgents2 moves drug discovery from single-task models to an end-to-end autonomous workflow: it distills transferable decision knowledge from published literature and uses parallel agents to explore target, hypothesis and strategy spaces. On the antibacterial target LpxH it autonomously hypothesized a previously unreported allosteric site and screened against it, yielding AIPher-L1 with experimentally confirmed inhibitory activity and binding; JAK1 work and secondary-target exploration for the known molecule AIPher-X were supported by cell assays and SPR experiments respectively.
PharmAgents2 moves drug discovery from single-task models to an end-to-end autonomous workflow: it distills transferable decision knowledge from published literature and uses parallel agents to explore target, hypothesis and strategy spaces. On the antibacterial target LpxH it autonomously hypothesized a previously unreported allosteric site and screened against it, yielding AIPher-L1 with experimentally confirmed inhibitory activity and binding; JAK1 work and secondary-target exploration for the known molecule AIPher-X were supported by cell assays and SPR experiments respectively.
Haichuan Tan（First author）
PharmAgents2 Paper2Skill Self-Scaling Discovery PharmAgents-Bench v0.1 Drug Discovery
望石智慧（StoneWise） Nov 11, 2026
StoneWise's MolVortex early-discovery agent lands in Lianyungang, opening to local pharma and academia via an AI tool marketplace
StoneWise's small-molecule early-discovery agent MolVortex has completed local deployment on Lianyungang's "AI + medicine" public service platform and been listed in its AI tool marketplace, open to local pharma companies, hospital researchers, universities and innovation service providers. The platform, invested in and built by Jiangsu XinHailian Group, went live at the 8th Lianyungang Pharmaceutical Technology Conference, aiming to let companies call AI capabilities and data assets on demand without building their own compute facilities. MolVortex covers target research, pocket analysis, molecule generation and dynamics evaluation across the early-discovery path from hit to preclinical candidate.
StoneWise's small-molecule early-discovery agent MolVortex has completed local deployment on Lianyungang's "AI + medicine" public service platform and been listed in its AI tool marketplace, open to local pharma companies, hospital researchers, universities and innovation service providers. The platform, invested in and built by Jiangsu XinHailian Group, went live at the 8th Lianyungang Pharmaceutical Technology Conference, aiming to let companies call AI capabilities and data assets on demand without building their own compute facilities. MolVortex covers target research, pocket analysis, molecule generation and dynamics evaluation across the early-discovery path from hit to preclinical candidate.
望石智慧（StoneWise）
MolVortex AI+Pharma public service platform cooperation Drug Discovery
arXiv Oct 17, 2026
An agent writes its execution experience into reviewable skills: DrugHarness designs and synthesizes an HPK1 compound with 14.19 nM cellular IC50
Computational drug-discovery agents usually rebuild their tool calls for every task, so a successful strategy rarely carries over. DrugHarness places more than 80 scientific operations and biomedical resources in one workspace, induces action hypotheses with explicit applicability conditions from execution trajectories, and admits them as skills only after an independent review agent checks sources and runs targeted trials when needed. On MolBench, the configuration without evolution reaches 91.9% affinity-comparison accuracy, 10.8 percentage points above reported MolClaw-CC results; in retrospective evaluation, frozen learned skills raise Qwen's optimization success from 79.5% to 87.2%.
Computational drug-discovery agents usually rebuild their tool calls for every task, so a successful strategy rarely carries over. DrugHarness places more than 80 scientific operations and biomedical resources in one workspace, induces action hypotheses with explicit applicability conditions from execution trajectories, and admits them as skills only after an independent review agent checks sources and runs targeted trials when needed. On MolBench, the configuration without evolution reaches 91.9% affinity-comparison accuracy, 10.8 percentage points above reported MolClaw-CC results; in retrospective evaluation, frozen learned skills raise Qwen's optimization success from 79.5% to 87.2%.
Ant Group (China)Nanjing University China Pharmaceutical University Shanghai Jiao Tong University Monash University Chinese Academy of Sciences Tongji University Westlake University
Chen Yuxing（First author）
DrugHarness Agent Harness Drug Discovery Skill Revision Lead Optimization
AstraZeneca Oct 15, 2026
AstraZeneca bets on insitro for MASH and cardiomyopathy: $40 million upfront to hunt causal targets with AI
AstraZeneca signed a new strategic drug discovery collaboration and licensing agreement with AI therapeutics company insitro to identify novel targets and advance new programmes for metabolic dysfunction-associated steatohepatitis (MASH) and cardiomyopathies. insitro will apply its Virtual Human AI platform, which combines large-scale human and cellular data with AI-enabled target identification and medicine design, to identify causal disease biology and move AstraZeneca-nominated targets from discovery to development candidate. insitro receives a $40 million upfront payment and is eligible for up to approximately $1.46 billion in development and commercial milestones plus single-digit royalties, while AstraZeneca holds exclusive worldwide rights to selected programmes.
AstraZeneca signed a new strategic drug discovery collaboration and licensing agreement with AI therapeutics company insitro to identify novel targets and advance new programmes for metabolic dysfunction-associated steatohepatitis (MASH) and cardiomyopathies. insitro will apply its Virtual Human AI platform, which combines large-scale human and cellular data with AI-enabled target identification and medicine design, to identify causal disease biology and move AstraZeneca-nominated targets from discovery to development candidate. insitro receives a $40 million upfront payment and is eligible for up to approximately $1.46 billion in development and commercial milestones plus single-digit royalties, while AstraZeneca holds exclusive worldwide rights to selected programmes.
AstraZeneca
Strategic Collaboration Drug Discovery AI-enabled target identification
Inductive Bio Sep 17, 2026
Beacon-2 turns human efficacious dose into a computable reward: agent cuts predicted dose 17-fold
Inductive released Beacon-2, a unified system that predicts a molecule's human efficacious dose end-to-end from chemical structure and treats that dose as the reward function for chemical optimization. AI drug discovery had focused largely on potency, one input to dose. On 325 compounds from the ExpansionRx blind challenge, fully blind dose predictions matched doses computed from measured ADME data with a Spearman rank correlation of 0.79 and an average fold error of 2.4x; across 20 partner programs the average rank correlation was 0.61.
Inductive released Beacon-2, a unified system that predicts a molecule's human efficacious dose end-to-end from chemical structure and treats that dose as the reward function for chemical optimization. AI drug discovery had focused largely on potency, one input to dose. On 325 compounds from the ExpansionRx blind challenge, fully blind dose predictions matched doses computed from measured ADME data with a Spearman rank correlation of 0.79 and an average fold error of 2.4x; across 20 partner programs the average rank correlation was 0.61.
Beacon-2 Efficacious Human Dose Prediction Drug Discovery
Nature Computational Science Oct 9, 2026
Sparse reaction data gets a closed loop: active learning picks the experiments, and geometric GNNs call every borylation site on unseen substrates
Late-stage functionalization in medicinal chemistry is limited by scarce reaction data, especially when several similar C–H bonds compete for borylation. A Roche and ETH team used an XGBoost-ensemble active learning loop to pick 50 previously unreported substrates from a 22,253-molecule pool across three rounds of high-throughput experiments, expanding C–H borylation data to 6,865 reactions over 568 substrates. Geometric graph neural networks trained on that set beat XGBoost in most cases on random, Butina and Bemis–Murcko scaffold splits, and in a prospective test on ten unseen substrates with N-heteroaryl motifs the top predicted site matched experiment in all cases.
Late-stage functionalization in medicinal chemistry is limited by scarce reaction data, especially when several similar C–H bonds compete for borylation. A Roche and ETH team used an XGBoost-ensemble active learning loop to pick 50 previously unreported substrates from a 22,253-molecule pool across three rounds of high-throughput experiments, expanding C–H borylation data to 6,865 reactions over 568 substrates. Geometric graph neural networks trained on that set beat XGBoost in most cases on random, Butina and Bemis–Murcko scaffold splits, and in a prospective test on ten unseen substrates with N-heteroaryl motifs the top predicted site matched experiment in all cases.
Roche F. Hoffmann-La Roche AG
Mason Minot（First author）Yannick Stenzhorn（First author）Gisbert Schneider（Corresponding author）David F. Nippa（Corresponding author）Kenneth Atz（Corresponding author）
Active Learning Geometric Deep Learning Regioselectivity Prediction C–H Borylation Drug Discovery
arXiv Oct 31, 2026
Risk-controlled candidate screening drops conformal p-values: UNEC calibrates the cutoff directly, taking about 1.3 seconds per repetition over the full simulation grid versus 35 to 37 for SCoRE
Machine learning screens far more candidates than can be experimentally validated, so selection must control risk before validation. UNEC normalizes predictive margins by local uncertainty and calibrates the selection cutoff directly on held-out losses, without building conformal p- or e-values, and supports general bounded losses and observation-specific thresholds. On solubility, toxicity and lipophilicity screening it holds risk-power trade-offs competitive with conformal selection, the NP method and SCoRE variants; in simulation the full grid took about 1.3 seconds per repetition versus 35 to 37 seconds for SCoRE.
Machine learning screens far more candidates than can be experimentally validated, so selection must control risk before validation. UNEC normalizes predictive margins by local uncertainty and calibrates the selection cutoff directly on held-out losses, without building conformal p- or e-values, and supports general bounded losses and observation-specific thresholds. On solubility, toxicity and lipophilicity screening it holds risk-power trade-offs competitive with conformal selection, the NP method and SCoRE variants; in simulation the full grid took about 1.3 seconds per repetition versus 35 to 37 seconds for SCoRE.
Stanford University McGill University Eli Lilly (United States)Tsinghua University
Ruinan Wang（First author）
Uncertainty-Normalized Empirical Calibration Risk-Controlled Selection Drug Discovery