Targeting BCL‑2 through Deep Learning-Based Drug Repurposing: A Multimodal Approach Combining Diffusion-Based Generative Modeling, Neural Relational Inference, and In Vitro Validation
Aug 2026· Journal of Chemical Information and Modeling· Vol 66, pp. 11361 - 11376· 0 citations· 35 references
Medicine
TL;DR
A multimodal BCL-2 repurposing workflow built with diffusion-based generative modeling for ligand-specific complex generation and an extended neural relational inference (NRI) framework for trajectory-level interaction analysis is presented, establishing a strategy that couples generative complex prediction with graph-based dynamic inference for structure-guided drug repurposing.
Abstract
Accurate identification of repurposable BCL-2 ligands requires not only plausible bound complex structures but also a dynamic description of how ligand binding reshapes residue-level communication. Here, we present a multimodal BCL-2 repurposing workflow built with diffusion-based generative modeling for ligand-specific complex generation and an extended neural relational inference (NRI) framework for trajectory-level interaction analysis. NeuralPlexer was applied to a library of 3094 FDA-approved drugs to generate BCL-2-ligand complex conformations at scale, yielding 1294 structurally acceptable complexes for downstream prioritization. To complement static scoring, filtered candidates were evaluated by molecular docking, anticancer QSAR classification, all-atom molecular dynamics (MD) simulations, and MM/GBSA binding free-energy calculations. We then extended NRI to protein–ligand trajectories to quantify residue-ligand and residue–residue dynamic couplings, enabling comparison of candidate-specific interaction signatures against the reference BCL-2 inhibitor Venetoclax. Among the prioritized compounds, Relugolix emerged as one of the most compelling hits, combining favorable binding energetics with an NRI-derived interaction pattern closely resembling that of Venetoclax. In vitro experiments supported BCL-2 inhibition by Relugolix in a TR-FRET assay and reduced viability of LN-18 glioma cells (IC50 = 23.55 μM). Together, these results establish a strategy that couples generative complex prediction with graph-based dynamic inference for structure-guided drug repurposing and identify Relugolix as a tractable scaffold for future BCL-2 inhibitor design.
This study applies MPS2IT-DTI (Molecule and Protein Sequence to Image Transformer for Drug-Target Interaction), a deep learning framework that represents molecular and protein sequences as images as images using k-mer frequency encoding, enabling convolutional neural networks to capture spatial compositional patterns associated with biochemical interactions.
J. D. de Souza, M. Fernandes, Raquel de Melo Barbosa· Computational biology and ch...· 0 citations
An integrated computational framework combining machine learning (ML), deep learning (DL), and structure-based docking with experimental validation identifies AO65 as a promising lead for further TDP1-focused investigation.
Huang Zeng, Man-Yi Zhang, Bo Qiu et al.· RSC Advances· 0 citations
Together, these results provide a structurally grounded inactive-state HER2 model and a mechanistic framework for structure-based design of HER2 inhibitors targeting inactive kinase conformations, highlighting the applicability of deep learning CVs to systems of complex free energy landscape.
Muhammad I. Ismail, Mai Adel, Eman M. E. Dokla et al.· Computational biology and ch...· 0 citations
These findings introduce H_1 as a computationally prioritized, putative MLK4-binding lead and provide a hypothesis-generating framework for MLK4-targeted scaffold prioritization, while recognizing that experimental activity and kinome selectivity profiling remain necessary before H_1 can be described as a confirmed MLK4 inhibitor or MLK4-selective compound.
Afnan A. Alzaghari, S. Daoud, Husam Nassar et al.· Journal of Pharmaceutical In...· 0 citations
Background/Objectives: Molecular docking and molecular dynamics are accurate but computationally expensive, so the compounds entering them must be chosen well. The present study proposes CADT, a confidence-gated affinity–ADME-T docking-triage cascade that decides which compounds are worth docking. Methods: The gate combines an ensemble estimate of drug–target affinity with its epistemic uncertainty and an applicability-domain check. Predicted absorption, distribution, metabolism, excretion, and toxicity (ADME-T) developability is added as a soft flag. All components were trained on openly licensed Therapeutics Data Commons data. Ranking was assessed on the DAVIS and KIBA kinase panels and on BindingDB Kd, under three split protocols over five seeds. The routing decision was then examined against molecular docking, in which 407 compound–target pairs were docked into six withheld kinases. Results: A Morgan-fingerprint gradient-boosting model reached a concordance index of 0.866±0.006, with 0.813 for unseen targets and 0.720 for unseen drugs. Across eight ADME-T endpoints, the area under the ROC curve ranged from 0.65 to 0.91. On the cold-target split the cascade reduced the compounds sent to docking by 86% while retaining 61% of the true strong binders. Docking measured that reduction at 85%, and at an equal budget, the gate enriched true binders more than the docking score itself. Conclusions: A transparent pre-screen can prioritise compounds ahead of structure-based calculation at a fraction of its cost. However, the uncertainty and applicability-domain terms act as an abstention mechanism rather than an accuracy gain, and that abstention is not free.
Drug-target affinity prediction provides a computational basis for virtual screening and drug repositioning optimization by estimating the binding strength between compounds and target proteins. Existing deep learning methods have evolved from early SMILES/amino acid sequence modeling to graph neural networks and multi-modal fusion. However, there are still three challenges: one-dimensional sequence models struggle to explicitly represent atomic connections and residue interactions; some graph models incorporate molecular or protein contact graphs, but often fail to properly align the semantics of drugs with those of proteins; some multi-modal approaches simply stack multiple source features together, without organizing the interaction processes based on different levels of binding information. To address these issues, this paper proposes MLI-DTA, a hierarchical multi-source feature fusion network based on multi-level attention. MLI-DTA encodes both sequence and graph modalities of drugs and targets. At a finer granularity level, the Bilinear Attention Network models high-level pairings between drug fragments and protein sequence fragments. At a medium granularity level, the Global Graph Cross-Attention mechanism aligns the topological structure of drugs with the contact graph of proteins. Additionally, the Multi-head Collaborative Attention mechanism integrates graph interaction vectors with sequence-level global representations. At a coarser granularity level, the Gate Fusion mechanism dynamically combines global features from both sequences and graphs. Multiple layers of information are combined to create an affinity predictor, enabling the model to utilize both sequence interactions, graph topological semantics, and global context simultaneously. Experiments conducted in the Davis and KIBA benchmarks demonstrate that MLI-DTA performs exceptionally well.
Yu-Ning Liu, Guang-Ze Wang, Dan Liu et al.· European journal of medicina...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
