Zhaoxi Sun

(+)-Magnolin is identified as a selective inhibitor for hydroxycarboxylic acid receptor 1 (HCAR1).

Precise regulation of chimeric antigen receptor (CAR)‐T cell activity is essential to maximize efficacy and minimize toxicity. While switch‐controlled CAR‐T holds great promise, challenges remain in achieving accurate activation, reducing immunogenicity, and preventing off‐target effects. Here we present a novel inducible ON‐switch CAR design that leverages plant hormone signaling components to achieve controllable T cell activation. By engineering a receptor system integrating the plant auxin receptor AFB1 with its co‐receptor IAA7, we enable ligand‐dependent interactions triggered by the plant hormone auxins. This design allows rapid, reversible, and dose‐dependent T cell activation, resulting in potent cytotoxicity against B‐cell lymphoma in vitro and in vivo. Notably, auxCAR‐T cells controlled by synthetic auxin analogs maintain a favorable memory phenotype and exhibit reduced functional exhaustion during treatment, leading to improved therapeutic efficacy. Our plant hormone‐based orthogonal system overcomes key limitations of existing switch systems and advances the development of precision CAR‐T therapies.

Computational blind challenges offer critical, unbiased opportunities to assess and accelerate scientific progress, as demonstrated by a breadth of breakthroughs over the past decade. We report the outcomes and key insights from an open science community blind challenge focused on computational methods in drug discovery, using lead optimization data from the AI-driven Structure-enabled Antiviral Platform Discovery Consortium's pan-coronavirus antiviral discovery program, in partnership with Polaris and the OpenADMET project. This collaborative initiative invited global participants from both academia and industry to develop and apply computational methods to predict the biochemical potency and crystallographic ligand poses of small molecules against key coronavirus targets, Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) and Middle East Respiratory Syndrome Coronavirus (MERS-CoV) main protease (Mpro), as well as multiple ADMET assay end points, using previously undisclosed comprehensive experimental drug discovery data sets as benchmarks. By evaluating submissions across multiple tasks and compounds, we established performance leaderboards and conducted meta-analyses to assess methodological strengths, common pitfalls, and areas for improvement. This analysis provides a foundation for best practices in real-world machine learning evaluation, grounded in community-driven benchmarking. We also highlight how next-generation platforms, such as Polaris, enable rigorous challenge design, embedded evaluation frameworks, and broad community engagement. This paper reports the collective findings of the challenge, offering a high-level overview of the data, evaluation infrastructure, and top-performing strategies. We further provide context and support for the accompanying papers authored by the challenge participants in this special issue, which explore individual approaches in greater depth. Together, these contributions aim to advance reproducible, trustworthy, and high-impact computational methods in drug discovery, and to explore best practices and pitfalls in future blind challenge design and execution, including planned initiatives for the OpenADMET project.

Targeted protein degradation (TPD), including PROTACs and molecular glues (MG), offers a path to modulate previously “undruggable” targets, but the multicomponent nature of MG systems complicates binding affinity evaluation, especially without crystal structures. We present a workflow integrating AI-assisted prediction with physics-driven free energy calculations. Across 28 crystal structures, AI-assisted structure prediction models perform well in reconstructing quaternary complexes, despite the fact that all tested entries are deposited after the proposal of the AI tool (i.e., not in the training set). For free energy estimation, instead of using individual scoring regimes, we propose a combined ASGBIE [ ] Vinardo strategy that significantly outperforms either method alone. The proposed approach enables the estimation of the bioactivity of MG from only sequence and ligand input, enabling a practical virtual screening. Further, energy decomposition enables the elucidation of protein–MG interaction mechanisms and highlights [ ]-cation and [ ]-methyl interactions as key contributors to MG-induced cooperativity. This integrated workflow provides mechanistic insight and practical guidance for MG design, supporting in silico discovery when experimental structures are unavailable.

Targeted covalent inhibitors (TCIs) have become an important modality in modern drug discovery, but computational tools for covalent pose prediction and quantitative affinity ranking remain underdeveloped. We constructed a large, structure and activity-resolved benchmark to systematically evaluate covalent docking and to develop a covalent-aware drug-target affinity (DTA) prediction framework. Starting from CovalentInDB 2.0 and related structural resources, we curated 2172 high quality covalent protein–ligand complexes spanning diverse protein classes and nine electrophilic warhead types, and used them to benchmark four docking engines (AutoDock4, CovDock in the Schrödinger Suite, GNINA and Boltz-2) in a self-docking setting. Boltz-2 shows the strongest pose-reproduction performance on our structure-resolved benchmark. However, because co-folding engines are trained on broad PDB corpora and our benchmark is also derived from PDB-resolved complexes, potential train-test overlap is likely; thus, Boltz-2 results are reported as a reference upper bound rather than a leakage-free estimate of prospective generalization. Across 17 covalent targets with quantitative IC₅₀ data, we further assessed the relationship between docking scores and experimental pIC₅₀ values and found that score-affinity correlations are generally weak and highly target dependent, with |r| < 0.2 for most target-software pairs and even pronounced negative correlations for several systems. We propose CovMTL-DTA to overcome these limitations, a covalent-aware multi-task DTA model that integrates ligand molecular graphs augmented with SMARTS-based warhead descriptors, pretrained protein sequence embeddings, cross-modal ligand–protein attention, and a task-relation module for inter-target transfer. Trained on curated covalent ligand-target pairs, the model outperforms classical machine-learning regressors and state-of-the-art deep DTA baselines, achieving a Pearson correlation of ∼0.77 with reduced RMSE and MAE on an independent test set. In an EGFR-focused virtual screening of ∼14 000 Michael-acceptor-containing compounds, the model prioritizes three clinically relevant EGFR covalent inhibitors within the top 1% of the ranked library and identifies structurally novel, favorable physicochemical properties hits. Our benchmark and model highlight both the strengths and limitations of current covalent docking and demonstrate how covalent-specific representations and multi-task learning can substantially improve affinity prediction and hit prioritization in covalent drug discovery.

Computational blind challenges offer critical, unbiased assessment opportunities to assess and accelerate scientific progress, as demonstrated by a breadth of breakthroughs over the last decade. We report the outcomes and key insights from an open science community blind challenge focused on computational methods in drug discovery, using lead optimization data from the AI-driven Structure-enabled Antiviral Platform (ASAP) Discovery Consortium’s pan-coronavirus antiviral discovery program, in partnership with Polaris and the OpenADMET project. This collaborative initiative invited global participants from both academia and industry to develop and apply computational methods to predict the biochemical potency and crystallographic ligand poses of small molecules against key coronavirus targets, Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) and Middle East Respiratory Syndrome Coronavirus (MERS-CoV) main protease (Mpro), as well as multiple ADMET assay endpoints, using previously undisclosed comprehensive experimental drug discovery datasets as benchmarks. By evaluating submissions across multiple tasks and compounds, we established performance leaderboards and conducted meta-analyses to assess methodological strengths, common pitfalls, and areas for improvement. This analysis provides a foundation for best practices in real-world machine learning evaluation, grounded in community-driven benchmarking. We also highlight how next-generation platforms, such as Polaris, enable rigorous challenge design, embedded evaluation frameworks, and broad community engagement. This paper reports the collective findings of the challenge, offering a high-level overview of the data, evaluation infrastructure, and top- performing strategies. We further provide context and support for the accompanying papers authored by the challenge participants in this special issue, which explore individual approaches in greater depth. Together, these contributions aim to advance reproducible, trustworthy, and high-impact computational methods in drug discovery, and to explore best practices and pitfalls in future blind challenge design and execution, including planned initiatives for the OpenADMET project.

We systematically dissect how host architecture, sampling strategy, and computational approximations govern the convergence and predictive accuracy of end-point binding free energies.

Cyclodextrins are amphiphilic macromolecular containers that are particularly valuable in applications ranging from biomedicine to environmental science. Despite years of development of computational techniques for cyclodextrin host‐guest coordination, accurate modelling of these supramolecular systems remains challenging. In this work, an integrative computational technique is presented to solve this problem. The protocol integrates a force‐field recalibration procedure, equilibrium enhanced sampling, nonequilibrium Hamiltonian switching, a work convolution algorithm, a series of finite‐size corrections, and energy decomposition analysis. A large‐scale survey of 222 CD host–guest systems is performed to demonstrate that the protocol enables the fast calculation of cyclodextrin host‐guest binding strength without compromise in accuracy, an unbiased capture of cyclodextrin dynamics in the free CD, and the multi‐modal behavior of the coordination patterns, and further the identification of the physicochemical driving force stabilizing host‐guest complexes. Especially, the protocol consistently provides accurate results for various systems while conventional transferable force fields systematically fail for larger and more flexible 𝛾‐CD. It thus opens new opportunities for high‐throughput screening and rational design across diverse macrocyclic host families.

Human serum albumin (HSA) as the most prevalent protein constitutes ~60% of the protein mass. In this work, we augment our previously released HSA database (HSADab) via the incorporation of AI-powered modelling. The constructed webserver www.hsadab.cn enables instant prediction of HSA binding affinities, plasma protein binding (PPB) and also their fragment-specific contributions for drug-like molecules through various machine-learning predictors, hosts the most comprehensive affinity and structure banks containing all HSA-relevant data published so far, and contains a complete set of deep-learning assisted docking structures for molecules presented in the database. We additionally present comprehensive analyses on the protein conformational space, docking performance and AlphaFold modelling. The updated database and the machine learning tools would accelerate drug discovery, especially in PK/PD process.