arXiv subjects

Luhua Lai

Publications and source records attributed to Luhua Lai.

14 recordsLinked to original sources

arXiv2026

Molexar: A Unified Multimodal Molecular Foundation Model for Drug Design

Molecular generation is a central challenge in drug discovery, requiring models that explore vast chemical space while satisfying diverse design constraints. We present Molexar, a unified multimodal molecular foundation model built on Fragment-SELFIES, a robust, fragment-aware molecular language with validity-preserving decoding and explicit fragment structure. A pretrained autoregressive decoder learns the Fragment-SELFIES syntax and molecular distribution; supervised fine-tuning (SFT) then trains the same decoder on condition-molecule pairs spanning scalar molecular properties, pharmacophore fingerprints, protein sequences, and binding pockets, injecting each condition by in-place replacement of value-token embeddings so that all generation modes share one autoregressive path. Molexar achieves strong efficiency at a small parameter count while matching or exceeding larger models. The pretrained model reaches 100% validity and high drug-likeness in unconditional and fragment-constrained generation; the SFT model follows single- and multi-property instructions and remains competitive on target-conditioned generation on the CrossDocked2020 test set. On MolGenBench, Molexar further generates molecules with favorable safety and potency. These results establish Molexar as a practical unified foundation for computational chemistry and drug-design workflows.

q-bio.BM↗

arXiv2025

SynCraft: Guiding Large Language Models to Predict Edit Sequences for Molecular Synthesizability Optimization

Generative artificial intelligence has revolutionized the exploration of chemical space, yet a critical bottleneck remains that a substantial fraction of generated molecules is synthetically inaccessible. Current solutions, such as post-hoc filtering or projection-based methods, often compromise structural novelty or disrupt key pharmacophores by forcing molecules into pre-defined synthetic templates. Herein, we introduce SynCraft, a reasoning-based framework that reframes synthesizability optimization not as a sequence translation task, but as a precise structural editing problem. Leveraging the emergent reasoning capabilities of Large Language Models, SynCraft navigates the "synthesis cliff" where minimal structural modifications yield significant gains in synthetic feasibility. By predicting executable sequences of atom-level edits rather than generating SMILES strings directly, SynCraft circumvents the syntactic fragility of LLMs while harnessing their chemical intuition. Extensive benchmarks demonstrate that SynCraft outperforms state-of-the-art baselines in generating synthesizable analogs with high structural fidelity. Furthermore, through interaction-aware prompting, SynCraft successfully replicates expert medicinal chemistry intuition in editing PLK1 inhibitors and rescuing high-scoring but previously discarded RIPK1 candidates in previous molecular generation literatures.

cs.AI↗

arXiv2025

GPx4 is bound to peroxidized membranes by a hydrophobic anchor

Ferroptosis is a form of cell death discovered in recent years, induced by excessive peroxidation of phospholipids. Glutathione peroxidase 4 (GPx4) is an intracellular enzyme that can repair the peroxidized phospholipids on membranes, thus regulating ferroptosis. By combining multiscale molecular dynamics (MD) simulations and experimental assays, we investigate the binding mechanisms of GPx4 on membranes. Using coarse-grained MD simulations, we found that L130 and its adjacent residues on GPx4 can form a stable and unique binding interface with PE/PS-rich and peroxidized membranes. Subsequent all-atom MD simulations verified the stability of the binding interface. The critical residue on the interface, L130, was inserted deeply into the membrane as a hydrophobic anchor and guided the reaction center toward the membrane surface. Enzyme activity assays and in vitro cell experiments showed that mutations of L130 resulted in weaker activities of the enzyme, probably caused by non-functional binding modes of GPx4 on membranes, as revealed by in silico simulations. This study highlights the crucial role of the hydrophobic residue, L130, in the proper anchoring of GPx4 on membranes, the first step of its membrane-repairing function.

q-bio.BM↗

arXiv2024

Target-Specific De Novo Peptide Binder Design with DiffPepBuilder

Despite the exciting progress in target-specific de novo protein binder design, peptide binder design remains challenging due to the flexibility of peptide structures and the scarcity of protein-peptide complex structure data. In this study, we curated a large synthetic dataset, referred to as PepPC-F, from the abundant protein-protein interface data and developed DiffPepBuilder, a de novo target-specific peptide binder generation method that utilizes an SE(3)-equivariant diffusion model trained on PepPC-F to co-design peptide sequences and structures. DiffPepBuilder also introduces disulfide bonds to stabilize the generated peptide structures. We tested DiffPepBuilder on 30 experimentally verified strong peptide binders with available protein-peptide complex structures. DiffPepBuilder was able to effectively recall the native structures and sequences of the peptide ligands and to generate novel peptide binders with improved binding free energy. We subsequently conducted de novo generation case studies on three targets. In both the regeneration test and case studies, DiffPepBuilder outperformed AfDesign and RFdiffusion coupled with ProteinMPNN, in terms of sequence and structure recall, interface quality, and structural diversity. Molecular dynamics simulations confirmed that the introduction of disulfide bonds enhanced the structural rigidity and binding performance of the generated peptides. As a general peptide binder de novo design tool, DiffPepBuilder can be used to design peptide binders for given protein targets with three dimensional and binding site information.

q-bio.BM↗

arXiv2024

DeepRLI: A Multi-objective Framework for Universal Protein--Ligand Interaction Prediction

Protein (receptor)--ligand interaction prediction is a critical component in computer-aided drug design, significantly influencing molecular docking and virtual screening processes. Despite the development of numerous scoring functions in recent years, particularly those employing machine learning, accurately and efficiently predicting binding affinities for protein--ligand complexes remains a formidable challenge. Most contemporary methods are tailored for specific tasks, such as binding affinity prediction, binding pose prediction, or virtual screening, often failing to encompass all aspects. In this study, we put forward DeepRLI, a novel protein--ligand interaction prediction architecture. It encodes each protein--ligand complex into a fully connected graph, retaining the integrity of the topological and spatial structure, and leverages the improved graph transformer layers with cosine envelope as the central module of the neural network, thus exhibiting superior scoring power. In order to equip the model to generalize to conformations beyond the confines of crystal structures and to adapt to molecular docking and virtual screening tasks, we propose a multi-objective strategy, that is, the model outputs three scores for scoring and ranking, docking, and screening, and the training process optimizes these three objectives simultaneously. For the latter two objectives, we augment the dataset through a docking procedure, incorporate suitable physics-informed blocks and employ an effective contrastive learning approach. Eventually, our model manifests a balanced performance across scoring, ranking, docking, and screening, thereby demonstrating its ability to handle a range of tasks. Overall, this research contributes a multi-objective framework for universal protein--ligand interaction prediction, augmenting the landscape of structure-based drug design.

q-bio.BM↗

arXiv2024

Accelerating Discovery of Novel and Bioactive Ligands With Pharmacophore-Informed Generative Models

Deep generative models have gained significant advancements to accelerate drug discovery by generating bioactive chemicals against desired targets. Nevertheless, most generated compounds that have been validated for potent bioactivity often exhibit structural novelty levels that fall short of satisfaction, thereby providing limited inspiration to human medicinal chemists. The challenge faced by generative models lies in their ability to produce compounds that are both bioactive and novel, rather than merely making minor modifications to known actives present in the training set. Recognizing the utility of pharmacophores in facilitating scaffold hopping, we developed TransPharmer, an innovative generative model that integrates ligand-based interpretable pharmacophore fingerprints with generative pre-training transformer (GPT) for de novo molecule generation. TransPharmer demonstrates superior performance across tasks involving unconditioned distribution learning, de novo generation and scaffold elaboration under pharmacophoric constraints. Its distinct exploration mode within the local chemical space renders it particularly useful for scaffold hopping, producing compounds that are structurally novel while pharmaceutically related. The efficacy of TransPharmer is validated through two case studies involving the dopamine receptor D2 (DRD2) and polo-like kinase 1 (PLK1). Notably in the case of PLK1, three out of four synthesized designed compounds exhibit submicromolar activities, with the most potent one, IIP0943, demonstrating a potency of 5.1 nM. Featuring a new scaffold of 4-(benzo[b]thiophen-7-yloxy)pyrimidine, IIP0943 also exhibits high selectivity for PLK1. It was demonstrated that TransPharmer is a powerful tool for discovery of novel and bioactive ligands.

q-bio.QM↗

arXiv2023

DiffBindFR: An SE(3) Equivariant Network for Flexible Protein-Ligand Docking

Molecular docking, a key technique in structure-based drug design, plays pivotal roles in protein-ligand interaction modeling, hit identification and optimization, in which accurate prediction of protein-ligand binding mode is essential. Conventional docking approaches perform well in redocking tasks with known protein binding pocket conformation in the complex state. However, in real-world docking scenario without knowing the protein binding conformation for a new ligand, accurately modeling the binding complex structure remains challenging as flexible docking is computationally expensive and inaccurate. Typical deep learning-based docking methods do not explicitly consider protein side chain conformations and fail to ensure the physical plausibility and detailed atomic interactions. In this study, we present DiffBindFR, a full-atom diffusion-based flexible docking model that operates over the product space of ligand overall movements and flexibility and pocket side chain torsion changes. We show that DiffBindFR has higher accuracy in producing native-like binding structures with physically plausible and detailed interactions than available docking methods. Furthermore, in the Apo and AlphaFold2 modeled structures, DiffBindFR demonstrates superior advantages in accurate ligand binding pose and protein binding conformation prediction, making it suitable for Apo and AlphaFold2 structure-based drug design. DiffBindFR provides a powerful flexible docking tool for modeling accurate protein-ligand binding structures.

q-bio.BM↗

arXiv2022

Synthesis-driven design of 3D molecules for structure-based drug discovery using geometric transformers

Finding drug-like compounds with high bioactivity is essential for drug discovery, but the task is complicated by the high cost of chemical synthesis and validation. With their outstanding performance in de novo drug design, deep generative models represent promising tools for tackling this challenge. In recently years, 3D molecule generative models have gained increasing attention due to their ability to directly utilize the 3D interaction information between the target and ligand. However, it remains challenging to synthesize the molecules generated by these models, limiting the speed of bioactivity validation and further structure optimization. In this work, we propose DeepLigBuilder+, a deep generative model for 3D molecules that combines structure-based de novo drug design with a reaction-based generation framework. Besides producing 3D molecular structures, the model also proposes synthetic pathways for generated molecules, which greatly assists the retro-synthetic analysis. To achieve this, we developed a new way to enforce the synthesizability constraint using a tree-based organization of purchasable building blocks. This method enjoys high scalability and is compatible with existing atom-based generative models. Additionally, for structure-based design tasks, we developed an SE(3)-equivariant transformer conditioned on the shape and pharmacophore-based inputs, and combine it with the Monte Carlo tree search. Using the ATP-binding pocket of BTK and the NAD+ binding pocket of PHGDH for case studies, we demonstrate that DeepLigBuilder+ is capable of enriching drug-like molecules with high predicted binding affinity and desirable interaction modes while maintaining the synthesizability constraint. We believe that DeepLigBuilder+ is a powerful tool for accelerating the process of drug discovery.

q-bio.QM↗

arXiv2022

MolMiner: You only look once for chemical structure recognition

Molecular structures are always depicted as 2D printed form in scientific documents like journal papers and patents. However, these 2D depictions are not machine-readable. Due to a backlog of decades and an increasing amount of these printed literature, there is a high demand for the translation of printed depictions into machine-readable formats, which is known as Optical Chemical Structure Recognition (OCSR). Most OCSR systems developed over the last three decades follow a rule-based approach where the key step of vectorization of the depiction is based on the interpretation of vectors and nodes as bonds and atoms. Here, we present a practical software MolMiner, which is primarily built up using deep neural networks originally developed for semantic segmentation and object detection to recognize atom and bond elements from documents. These recognized elements can be easily connected as a molecular graph with distance-based construction algorithm. We carefully evaluate our software on four benchmark datasets with the state-of-the-art performance. Various real application scenarios are also tested, yielding satisfactory outcomes. The free download links of Mac and Windows versions are available: Mac: https://molminer-cdn.iipharma.cn/pharma-mind/artifact/latest/mac/PharmaMind-mac-latest-setup.dmg and Windows: https://molminer-cdn.iipharma.cn/pharma-mind/artifact/latest/win/PharmaMind-win-latest-setup.exe

cs.CV↗

arXiv2021

Learning to design drug-like molecules in three-dimensional space using deep generative models

Recently, deep generative models for molecular graphs are gaining more and more attention in the field of de novo drug design. A variety of models have been developed to generate topological structures of drug-like molecules, but explorations in generating three-dimensional structures are still limited. Existing methods have either focused on low molecular weight compounds without considering drug-likeness or generate 3D structures indirectly using atom density maps. In this work, we introduce Ligand Neural Network (L-Net), a novel graph generative model for designing drug-like molecules with high-quality 3D structures. L-Net directly outputs the topological and 3D structure of molecules (including hydrogen atoms), without the need for additional atom placement or bond order inference algorithm. The architecture of L-Net is specifically optimized for drug-like molecules, and a set of metrics is assembled to comprehensively evaluate its performance. The results show that L-Net is capable of generating chemically correct, conformationally valid, and highly druglike molecules. Finally, to demonstrate its potential in structure-based molecular design, we combine L-Net with MCTS and test its ability to generate potential inhibitors targeting ABL1 kinase.

q-bio.QM↗

arXiv2019

Automatic Retrosynthetic Pathway Planning Using Template-free Models

We present an attention-based Transformer model for automatic retrosynthesis route planning. Our approach starts from reactants prediction of single-step organic reactions for given products, followed by Monte Carlo tree search-based automatic retrosynthetic pathway prediction. Trained on two datasets from the United States patent literature, our models achieved a top-1 prediction accuracy of over 54.6% and 63.0% with more than 95% and 99.6% validity rate of SMILES, respectively, which is the best up to now to our knowledge. We also demonstrate the application potential of our model by successfully performing multi-step retrosynthetic route planning for four case products, i.e., antiseizure drug Rufinamide, a novel allosteric activator, an inhibitor of human acute-myeloid-leukemia cells and a complex intermediate of drug candidate. Further, by using heuristics Monte Carlo tree search, we achieved automatic retrosynthetic pathway searching and successfully reproduced published synthesis pathways. In summary, our model has achieved the state-of-the-art performance on single-step retrosynthetic prediction and provides a novel strategy for automatic retrosynthetic pathway planning.

q-bio.QM↗

arXiv2017

Deep Learning Based Regression and Multi-class Models for Acute Oral Toxicity Prediction with Automatic Chemical Feature Extraction

For quantitative structure-property relationship (QSPR) studies in chemoinformatics, it is important to get interpretable relationship between chemical properties and chemical features. However, the predictive power and interpretability of QSPR models are usually two different objectives that are difficult to achieve simultaneously. A deep learning architecture using molecular graph encoding convolutional neural networks (MGE-CNN) provided a universal strategy to construct interpretable QSPR models with high predictive power. Instead of using application-specific preset molecular descriptors or fingerprints, the models can be resolved using raw and pertinent features without manual intervention or selection. In this study, we developed acute oral toxicity (AOT) models of compounds using the MGE-CNN architecture as a case study. Three types of high-level predictive models: regression model (deepAOT-R), multi-classification model (deepAOT-C) and multi-task model (deepAOT-CR) for AOT evaluation were constructed. These models highly outperformed previously reported models. For the two external datasets containing 1673 (test set I) and 375 (test set II) compounds, the R2 and mean absolute error (MAE) of deepAOT-R on the test set I were 0.864 and 0.195, and the prediction accuracy of deepAOT-C was 95.5% and 96.3% on the test set I and II, respectively. The two external prediction accuracy of deepAOT-CR is 95.0% and 94.1%, while the R2 and MAE are 0.861 and 0.204 for test set I, respectively.

stat.ML↗

arXiv2006

Robustness and modular design of the Drosophila segment polarity network

Biomolecular networks have to perform their functions robustly. A robust function may have preferences in the topological structures of the underlying network. We carried out an exhaustive computational analysis on network topologies in relation to a patterning function in Drosophila embryogenesis. We found that while the vast majority of topologies can either not perform the required function or only do so very fragilely, a small fraction of topologies emerges as particularly robust for the function. The topology adopted by Drosophila, that of the segment polarity network, is a top ranking one among all topologies with no direct autoregulation. Furthermore, we found that all robust topologies are modular--each being a combination of three kinds of modules. These modules can be traced back to three sub-functions of the patterning function and their combinations provide a combinatorial variability for the robust topologies. Our results suggest that the requirement of functional robustness drastically reduces the choices of viable topology to a limited set of modular combinations among which nature optimizes its choice under evolutionary and other biological constraints.

q-bio.MN↗

arXiv2005

Specificity of Trypsin and Chymotrypsin: Loop Motion Controlled Dynamic Correlation as a Determinant

Trypsin and chymotrypsin are both serine proteases with high sequence and structural similarities, but with different substrate specificity. Previous experiments have demonstrated the critical role of the two loops outside the binding pocket in controlling the specificity of the two enzymes. To understand the mechanism of such a control of specificity by distant loops, we have used the Gaussian Network Model to study the dynamic properties of trypsin and chymotrypsin and the roles played by the two loops. A clustering method was introduced to analyze the correlated motions of residues. We have found that trypsin and chymotrypsin have distinct dynamic signatures in the two loop regions which are in turn highly correlated with motions of certain residues in the binding pockets. Interestingly, replacing the two loops of trypsin with those of chymotrypsin changes the motion style of trypsin to chymotrypsin-like, whereas the same experimental replacement was shown necessary to make trypsin have chymotrypsin's enzyme specificity and activity. These results suggest that the cooperative motions of the two loops and the substrate-binding sites contribute to the activity and substrate specificity of trypsin and chymotrypsin.

q-bio.BM↗