Biology subjects

Xi, B.

Publications and source records attributed to Xi, B..

7 recordsLinked to original sources

bioRxiv2026

A closed-loop reinforcement learning framework for rapid compound directed optimization

Generative artificial intelligence (AI) holds transformative potential for drug discovery, yet existing architectures typically operate in open loops without experimental feedback. Here we introduce rapid compound directed optimization (RCDO), a closed-loop reinforcement learning framework that accelerates the optimization process by bridging dry-lab computation with wet-lab feedback. RCDO couples a three-dimensional structure-guided generative model with a multi-level reward system updated after each design cycle using experimental measurements from all synthesized compounds, including inactive or developability-failed compounds. By continuously aligning the generative model with accumulated wet-lab measurements, RCDO substantially compresses optimization timelines. We evaluated RCDO through retrospective benchmarking against historical optimization trajectories and prospective wet-lab campaigns targeting ROR1, NLRP3, and NSD3. Across prospective evaluations, RCDO rapidly resolved key optimization bottlenecks within two to three design cycles: improving the oral exposure of an ROR1 inhibitor by 40-fold while maintaining antitumor efficacy, reducing CYP2C19 inhibition of an NLRP3 antagonist by 20-fold while preserving inflammasome activity, and boosting the binding affinity of an NSD3 hit by 18-fold. By directly coupling wet-lab feedback to generative learning, RCDO establishes an efficient platform for compound directed optimization, transforming AI-driven drug discovery from static generation into continuous experimental adaptation.

bioinformatics↗

bioRxiv2025

A unified language model bridging de novo and fragment-based 3D molecule design delivers potent CBL-B inhibitors for cancer treatment

The rational design of small molecules is central to drug discovery, yet current artificial intelligence (AI) methodologies for generating three-dimensional (3D) molecules are often siloed, focusing on either de novo design or fragment-based design. The lack of a holistic framework limits AIs application across the complex and multi-step pipeline spanning from novel scaffold identification to lead compound optimization, and prevents AI from effectively learning from the entire process. Here, we introduce UniLingo3DMol, a language model for 3D molecular generation, empowered by fragment permutation-capable molecular representation alongside multi-stage and multi-task training strategy. This integrated design enables UniLingo3DMol to seamlessly span both de novo and fragment-retained molecular design, demonstrating superior performance over existing generation models in in silico evaluations across more than 100 diverse biological targets. We further leveraged UniLingo3DMol in the design of inhibitors targeting CBL-B, a crucial immune E3 ubiquitin ligase and attractive immunotherapy target. This strategy led to a lead compound demonstrating excellent in vitro activity and robust in vivo anti-tumor efficacy. Our findings establish UniLingo3DMol as a generalized and powerful platform, showing the strong potential to advance AI-driven drug discovery.

bioinformatics↗

bioRxiv2025

Pocket-based molecule generation with an SE(3)-equivariant language model leads to a potent and selective HPK1 inhibitor with in vivo efficacy

Deep learning shows promise in structure-based drug discovery, yet challenges persist in generating pharmacologically plausible molecules with valid 3D conformation and decent binding mode in the pocket. We introduce SE3-BiLingoMol, an SE(3)-equivariant Transformer for pocket-based 3D molecule generation, addressing two key limitations of existing language-model approaches. First, it uses Geometric Algebra Transformers for SE(3)-equivariant handling of continuous 3D coordinates. Second, a bidirectional attention mechanism mitigates conformational errors accumulated during autoregressive sampling. These innovations enable SE3-BiLingoMol to generate 2D drug-like, 3D geometrically valid molecules with superior binding modes. Validated on DUD-E dataset containing over 100 targets, the model achieved state-of-the-art performance in de novo design and optimization. We applied SE3-BiLingoMol to design potent and selective inhibitors for HPK1, a promising immunotherapy target. Through an iterative human-AI workflow, integrating AI generation with experimental validations (X-ray crystallography, bioassays), we identified Cmpd. 6. This novel tetracyclic compound demonstrates potent HPK1 inhibition, excellent cellular activity, favorable pharmacokinetics, and robust anti-tumor in vivo efficacy as monotherapy and with PD-1 blockade. Our work establishes a sophisticated generative AI framework for 3D molecule design and demonstrates its application in developing cancer immunotherapy. Highlights[bullet] SE(3)-Equivariant Language Model for Molecule Generation: We introduce a dual-channel SE(3)-equivariant language model that supports de novo molecule design and optimization, addressing key challenges in prior language model-based 3D molecule generation approaches. [bullet]Discovery of Potent HPK1 Inhibitors: Application of this model led to the discovery of highly selective HPK1 inhibitors with robust in vivo efficacy, resulting in a promising lead compound. [bullet]AI-Guided Rational Drug Development Paradigm: This work establishes a paradigm for AI-guided rational drug development, showcasing AI as a "co-pilot" that inspires drug designers with novel ideas in an "evidence-based context".

bioinformatics↗

bioRxiv2024

Assessing Conformation Validity and Rationality of Deep Learning-Generated 3D Molecules

Recent advancements in artificial intelligence (AI) have revolutionized the field of 3D molecule generation. However, the lack of effective evaluation methods for 3D conformations limits further improvements. Current techniques, in order to achieve the necessary speed for evaluating large numbers of AI-generated molecules, often rely on empirical geometric metrics that do not adequately capture various conformational anomalies, or on molecular mechanics (MM) energy metrics that exhibit low accuracy and lack atomic or torsional details. To address this gap, we propose a two-stage approach that achieves both high speed and quantum mechanical (QM) level accuracy. The first stage, termed the validity test, employs an AI-derived force field to identify atoms with elevated energy resulting from implausible neighboring environments. The second stage, known as the rationality test, utilizes a deep learning network trained on data with density functional theory (DFT) accuracy to detect rotatable bonds with high torsional energies. To demonstrate the functionality of our evaluation system, we applied our approach to five recently reported 3D molecule generation AI models across 102 targets in Directory of Useful Decoys-Enhanced (DUD-E) dataset. To facilitate accessibility for the academic community, our method is available as an open-source package.

bioinformatics↗

bioRxiv2021

AutoVEM2: a flexible automated tool to analyze candidate key mutations and epidemic trends for virus

In our previous work, we developed an automated tool, AutoVEM, for real-time monitoring the candidate key mutations and epidemic trends of SARS-CoV-2. In this research, we further developed AutoVEM into AutoVEM2. AutoVEM2 is composed of three modules, including call module, analysis module, and plot module, which can be used modularly or as a whole for any virus, as long as the corresponding reference genome is provided. Therefore, its much more flexible than AutoVEM. Here, we analyzed three existing viruses by AutoVEM2, including SARS-CoV-2, HBV and HPV-16, to show the functions, effectiveness and flexibility of AutoVEM2. We found that the N501Y locus was almost completely linked to the other 16 loci in SARS-CoV-2 genomes from the UK and Europe. Among the 17 loci, 5 loci were on the S protein and all of the five mutations cause amino acid changes, which may influence the epidemic traits of SARS-CoV-2. And some candidate key mutations of HBV and HPV-16, including T350G of HPV-16 and C659T of HBV, were detected. In brief, we developed a flexible automated tool to analyze candidate key mutations and epidemic trends for any virus, which would become a standard process for virus analysis based on genome sequences in the future. HighlightsO_LIAn automatic tool to quickly analyze candidate key mutations and epidemic trends for any virus was developed. C_LIO_LIOur integrated analysis method and tool could become a standard process for virus mutation and epidemic trend analysis based on genome sequences in the future. C_LIO_LIN501Y with the other 16 highly linked mutation sites of SARS-CoV-2 in the UK and Europe were further confirmed, and some valuable mutation sites of HBV and HPV-16 were detected. C_LI

genomics↗

bioRxiv2021

Molecular dynamics simulation study of effects of key mutations in SARS-CoV-2 on protein structures

SARS-CoV-2 has been spreading rapidly since 2019 and has produced large-scale mutations in the genomes. The mutation in genes may lead to changes in protein structure, which would have a great impact on the epidemiological characteristics. In this study, we selected the key mutations of SARS-CoV-2 from a real-time monitoring tool, including D614G, A222V, N501Y, T716I, S982A, D1118H of spike (S) protein, and performed molecular dynamics (MD) simulations on single-site mutant D614G, double-site mutant D614G&A222V and penta-site mutant N501Y&D614G&T716I&S982A&D1118H to investigate their effects on protein structure and stability using molecular dynamics (MD) simulations. The results suggested that D614G improved the stability of S protein, while D614G&A222V and N501Y&D614G&T716I&S982A&D1118H showed an increased solvent accessible surface area and they might enhance the ability of protein to react with the outside environment. Our findings could complement the mechanistic link between genotype--phenotype--epidemiological characteristics in the study of SARS-CoV-2. We also found no significant difference between the antigenicity of S protein and the mutants through Ellipro, which may reference for vaccine development and application.

bioinformatics↗

bioRxiv2020

Real-time monitoring epidemic trends and key mutations in SARS-CoV-2 evolution by an automated tool

With the global epidemic of SARS-CoV-2, it is important to monitor the variation, haplotype subgroup epidemic trends and key mutations of SARS-CoV-2 over time effectively, which is of great significance to the development of new vaccines, the update of therapeutic drugs, and the improvement of detection reagents. The AutoVEM tool developed in the present study could complete all mutations detections, haplotypes classification, haplotype subgroup epidemic trends and key mutations analysis for 131,576 SARS-CoV-2 genome sequences in 18 hours on a 1 core CPU and 2G internal storage computer. Through haplotype subgroup epidemic trends analysis of 131,576 genome sequences, the great significance of the previous 4 specific sites (C241T, C3037T, C14408T and A23403G) was further revealed, and 6 new mutation sites of highly linked (T445C, C6286T, C22227T, G25563T, C26801G and G29645T) were discovered for the first time that might be related to the infectivity, pathogenicity or host adaptability of SARS-CoV-2. In brief, we proposed an integrative method and developed an efficient automated tool to monitor haplotype subgroup epidemic trends and screen out the key mutations in the evolution of SARS-CoV-2 over time for the first time, and all data could be updated quickly to track the prevalence of previous key mutations and new key mutations because of high efficiency of the tool. In addition, the idea of combinatorial analysis in the present study can also provide a reference for the mutation monitoring of other viruses.

genomics↗