This Week In Cheminformatics: Issue #029

McKenna, Šícho, and colleagues built COMBINAUT, a repurposed peptide synthesizer running SPS workflows (amide/sulfonamide coupling, reductive amination, SNAr, nitro reduction, heterocyclization) across 715 in-house building blocks, enumerating a 22.9M compound virtual library and physically producing hits within 32h instead of the 4-6 week make-on-demand turnaround !! They report that not all but 78/100 selected VS hits actually got synthesized to spec, and only 9 of those were validated in radioligand binding against the CCR2 allosteric pocket (12% hit rate, pretty good !). They also try to justify these numbers in the paper carefully. They then used the same platform for rapid SAR on two hits, getting 4-6x potency improvements in a few hours of parallel synthesis per round, ending with two cell-active antagonists (IC50 256 nM and 1.3 µM in β-arrestin assays). It’s a modest library by ultra-large VS standards (22.9M vs billions), but the closed-loop docking followed by synthesis, assay, re-docking cycle all in hours rather than weeks is the actual contribution here, and it’s pretty amazing if you think about it.

M-JEPA ports the JEPA recipe (context encoder, EMA teacher, latent-space prediction instead of contrastive InfoNCE) to molecular graphs, using BFS-grown connected-subgraph masks as the target region. Under matched compute, JEPA beats a matched InfoNCE baseline on ESOL proxy RMSE and trains faster, and hybrid fine-tuning from the JEPA checkpoint lifts Tox21 mean ROC-AUC only modestly (0.561 to 0.609) but drops mean ECE substantially (0.279 to 0.064) which is a useful reminder that ranking and calibration under scaffold “shift” are separable properties and gains in one don’t guarantee the other. Their connected vs noncontiguous masking ablation was statistically non-signigicant on both proxy RMSE and motif-level coherence, so the “chemically coherent” argument for BFS masking seems basically aesthetic.

Here, Fite & Gross add a neat methodological contribution to the reaction network reduction problem. Most reduction algorithms are validated on a handful of published mechanisms, which tells you little about generality. The authors build a preferential attachment based generator fitted to ammonia, methane, and hydrogen combustion mechanisms, producing 2,613 artificial networks with sampled thermodynamics and Arrhenius kinetics, to get a large, unbiased test set. On top of that they introduce MolRank, a PageRank style topological algorithm that estimates upper bounds on maximal species concentration and reaction rate using only thermodynamic data. MolRank is markedly better at identifying redundant species than redundant reactions. They also report Eyring-Polanyi based scores give better bounds than true rate constants but don’t preserve ranking quality.

Klein et al. (BioSolveIT + Pfizer) benchmark SpaceMACS, SpaceLight, and FTrees across 2,917 CHEMBL queries on REAL Space, eXplore, and PfizerVL, and the headline result is that the three methods barely overlap under 1.2% shared hits across all three for a given query. MCS similarity, fingerprint (ECFP4/fCSFP) similarity, and FTrees pharmacophore similarity are sampling quite different neighborhoods of a chemical space, so which one and at what thresholds should you use them becomes the real question. This comparison was done with similarity scored on synthons rather than fully enumerated products. New bond formed at assembly for instance are invisible to standard ECFP4, which is why fCSFP was built with smaller, non-circular features to survive the synthon boundary better (recovering 90-100% of close analogs vs 70-95% for ECFP4). This paper gives you calibrated per-algorithm recommended thresholds for very similar, similar, and dissimilar compounds from search results (check Figure 16). Good read.

• Computational discovery of DENV-3 RdRp allosteric inhibitors: biological evaluation and mechanistic studies

• Meta-learning GNN with MD-informed attention for cross-species prediction of phosphoinositide-dependent kinase-1 (PdK1) inhibitors in termite control

• DeepDPM: A Deep Learning Method for MoRFs Prediction Based on Wavelet Transform and Dynamic Convolutional Attention Mechanism

• Modeling the Sensitivity of Large-Scale Virtual Screening to Scoring Function Accuracy, Artifacts, and Library Composition

• Reaction Enumeration Based on NBO-Informed Molecular Graphs

• Transition Mechanism of the Human Dopamine Transporter via Molecular Dynamics Simulations

• Enhancing Generalization in Synthesizability Prediction of Structurally Dissimilar Materials

• Allosteric Inhibition Modulates SHP2 Diffusion Behavior from Rotational to Translational via C-SH2 Domain Stabilization

• DeepPepQSAR: all-in-one for comprehensively exploiting the vast molecular diversity space of bioactive peptide universe

• Chemical features consistently associated with kinase inhibitor activity: a cross-target SHAP analysis under scaffold-split evaluation

• Discovery of potent biphenyl-macolacin derivatives through a lysine-T/CDHA iterative scanning (LTIS) strategy followed by chemical ligation-based modifications

• Prediction and mechanistic insights into drug-induced reproductive toxicity through integrated machine learning, FAERS-based signal comparison, and network toxicology analyses

• Resolving Conformational Preferences of Monosaccharides from 1H and 13C NMR Chemical Shifts Using an Integrated MD and QM Approach

• EnsembleCycPerm: Interpretable Modeling of Cyclic Peptide Permeability through Solvent-Dependent Conformational Ensembles

• Circumventing the Synthesizability Problem in Generative Molecular Design

• Unveiling Large-Scale Kinase-Centric Protein–Protein Interactions through a Knowledge-Informed Workflow

• Quantum Computing Calculations of Protein–Ligand Binding Energies Using Decomposition Methods on Simulated and Real Quantum Hardware

• A Numerical Implementation to Calculate Elastic Properties of Biological Membrane Simulations

• Integrated deep learning and multi-scale modeling for the discovery of pan-genotypic HCV NS5B polymerase inhibitors

• Integrating lncRNA data for prediction of miRNA-disease association using network fusion and matrix completion

• Unraveling the Mechanism of Drug Binding to SARS-CoV-2 RNA Pseudoknot With Thermodynamics-Driven Machine Learning

• Navigating chemical-linguistic sharing space with heterogeneous molecular encoding

• Covalent phosphorylation of poly(vinyl alcohol) eliminates phosphoric acid leaching without reducing proton donor activity: a short communication

• The Chemical Smiler: CHEMICal Abstraction Leading to SMILEs of Reference

• Reinforcement Learning-Driven Multiproperty Optimization in Molecular Design Using Multicontext Transcriptome Data

• PSDTA: An Approach to Drug-Target Binding Affinity Prediction by Integrating Physicochemical and Structural Information to Reduce Feature Redundancy

• Structural Proteomics-Based Deciphering of Hydrophobic Packing Fingerprints Informing Protein Thermostability in TIM Barrels

• Mapping Evolution of Molecules across Biochemistry with Assembly Theory

• Bridging between Structure-Based and Data-Driven Affinity Prediction

• Machine Learning-Driven Discovery of Indole/Oxoindole–Piperazine Scaffolds as Dual MAO-B/Sig-1R Ligands for Neurodegenerative Disorders

• Molecular Mechanism of Chiral Recognition of R- and S-Methamphetamine by the Human Dopamine Transporter: Insights from Thermodynamic and Kinetic Analyses

• Heterogeneous dual-channel and interpretable graph representation learning with global virtual nodes for microRNA-mediated drug sensitivity prediction

• Tightly coupled equivariant flow matching for molecular docking with multimodal physical constraints

• OpenCafeMol With 3SPN.2 DNA Model: GPU Acceleration for Long-Time Coarse-Grained Chromatin Simulations

• SpaBiT: enhancing spatial transcriptomics resolution via bidirectional attention transformers

• IDR searcher: a search engine solution for public image resources

• Meta2DB: curated shotgun metagenomic feature sets and metadata for health state prediction

• Partial Least Squares Regression Models for Outdoor Air Pollutant Forecasting

• Guiding generative models to uncover diverse and novel crystals via reinforcement learning

• Molecular dynamics simulation of the atropine drug encapsulation inside protein origami icosahedral nanocages in physiological environment

• FLOWR.ROOT – A flow matching-based foundation model for joint multi-purpose structure-aware 3D ligand generation and affinity prediction

• ScrambleBench: a workflow for comparative assessment of structure-based de novo generative models

• MINERVA: a public XAI-powered platform advancing multi-target discovery in Alzheimer’s disease

• Smiles-based bioactivity prediction through molecular encoder selection and data augmentation

• Higher-Order Dynamic Disentangled Intent Sensing and Bidirectional Joint Updating Framework for NcRNA-Drug Resistance Association Prediction

• TPS-Flow: Physics-Guided Flow-Based Generative Modeling of Protein Transition Paths

• Collective Variable-Guided Engineering of the Free-Energy Surface of a Small Peptide

• Can LLMs Solve Solubility Tasks? The SoluBench Benchmark for Pure and Mixed Solvent Systems

• Deep Learning Models Capture Umbrella Sampling-Derived Energetic Trends: A Troponin C Case Study

• HTS-Oracle v2: Prospective Ai-Guided Discovery and Experimental Validation of Small Molecule Modulators Across Multiple Targets

• RetroMPA: A Molecular Property-Aware Auxiliary Framework for Enhancing Retrosynthesis Prediction

• Insights into the Indexing of Powder X-ray Diffraction from a Robust Transformer Deep Learning

• NeuroPpred-PSCG: A Multimodal Framework Using ProtT5 and Structural Features for Neuropeptide Prediction Based on Gated Cross-Attention Mechanism and Multiscale Gated Convolution

• The BOS-Lig Data Set: Accurate Ligand Charges from a Consensus Approach for 66,810 Experimentally Synthesized Ligands

• ORIGAMI: Orientation-Aware Graph Neural Network for Assessing Multimeric Interfaces of Protein Complex Structures

• Evo-EquiGPS: Synergizing Dynamic Geometry, Global Topology, and Explicit Evolution for High-Precision Enzyme Active Site Prediction

• Protein Language Model-Based Fitness Estimates Facilitate Resistance Mutation Identification

• Hybrid Solvation Model for Analyzing the Binding of Yttrium and Calcium Cations to the Lanmodulin Protein Using the Fragment Molecular Orbital Method

• Protein Frustration Reveals Orthosteric and Allosteric Active Sites in GPCR:G Protein Complexes

• ProphDR: An Interpretable Deep Learning Model for Predicting Cancer Drug Response via Multi-Omics and Cross-Attention Mechanisms

• Tetrapeptide inhibitors of BACE-1 revealed by combined data-driven screening and physics-based free-energy refinement

• Discovery of novel cinnamic acid derivatives with anti-Helicobacter pylori mechanism

• Two faces of drug repurposing data in machine learning-driven structure-based virtual screening supported by interaction fingerprints

• A linear models approach to optimize carbazole-based dyes for solar cell applications

• DOCKweb: a web-based GUI platform for molecular modeling with DOCK6

• Synthesis and biological evaluation of bisspirooxindole-hexahydroindolizines as potential anticancer agent

• From inhibition to degradation: advances in highly selective targeting strategies for HPK1

• Nitrogen atom insertion and related skeletal editing strategies in aromatic heterocycles for drug discovery (2015–2025)

• Targeted Degradation of KAT6A: Expanding the Therapeutic Frontier of Epigenetic Drug Discovery

• Bicyclo[1.1.1]pentane as a Phenyl Isosteric Replacement

• Riding toward Selectivity: Optimization of Covalent 7-Azaindole-Based BMX Kinase Inhibitors

Ray@raysan5

On this day, 13 years ago, I asked in official Khronos forums for a simple and easy-to-use C library to teach videogames programming. I got no answer, so I created #raylib 🔥 If AI/LLMs had existed back then, would raylib exist today? 🤔 Forum: community.khronos.org/t/graphics-h-e…

9:29 AM · Jul 10, 2026 · 27.1K Views

33 Replies · 51 Reposts · 1.13K Likes

𝔞𝔡𝔦𝔫𝔞@haunt0l0gy

getting some remarkable targeted ads lately

6:23 PM · Jul 10, 2026 · 1.57M Views

158 Replies · 4.87K Reposts · 106K Likes

Jake Wintermute 🧬/acc@SynBio1

The virgin frontier LLM: "I can't tell you what a gene is the infohazard risk is simply too high" The chad academic journal: "Here's a method for building 5000 virus genomes all at once"

Nature Biotechnology @NatureBiotech

High-quality phage assembly from metagenomes with PALACE https://t.co/Q08gA28eAS

9:36 PM · Jul 9, 2026 · 18.1K Views

3 Replies · 14 Reposts · 203 Likes

John Gallagher@MereSophistry

So…Claude can’t do PHD level biology.

Anirban Maitra @Aiims1742

Anthropic is hiring a Research Scientist (post PhD) with biology & computational skills in its life sciences division for a $300-$320 K salary (2-3 times what a starting Assistant Professor would make in academia, after a serious postdoctoral fellowship). Brave new world!

7:56 PM · Jul 10, 2026 · 23.2K Views

6 Replies · 7 Reposts · 158 Likes

Pukicho@pukicho

Whats the point of a camera that lies to you

由依🍥 @youy1qwq

华为长焦拍飞机被识别成鸟了

9:01 PM · Jul 10, 2026 · 1.52M Views

196 Replies · 3.19K Reposts · 94.4K Likes

Saurabh Kumar@drummatick

I miss those days when programming discussions were fun

6:35 AM · Jul 10, 2026 · 254K Views

27 Replies · 138 Reposts · 7.2K Likes

Mathieu@miniapeur

10:09 AM · Jul 10, 2026 · 87K Views

18 Replies · 269 Reposts · 3.14K Likes

Ash Jogalekar@curiouswavefn

Freeman Dyson on attending a talk by Ed Witten in 1985 (Infinite in all Directions).

4:15 AM · Jul 11, 2026 · 11.6K Views

1 Reply · 32 Reposts · 223 Likes

Anshul Kundaje@anshulkundaje

Here is the recording of the talk and the paper making a case for why temporal dynamic data & sequence anchored cis-regulation will be critical to learn causal mechanistic insights into transcriptional regulation of perturbation response. 1/ https://t.co/9A0HN1bixm

Anshul Kundaje @anshulkundaje

Virtual cell enthusiasts: check out my talk today (in an hour) to understand why I think it is critical to have longitudinal (temporal) data & incorporate cis regulation into causal mechanistic models of perturbation response. Case study: fibroblast to iPSC reprogramming.

7:51 PM · Jul 10, 2026 · 25.6K Views

1 Reply · 16 Reposts · 153 Likes

Jason Burne@Monty_Brogan69

Au moins on se bidonait à l'époque avec le cyclisme full dopage assumé.

4:14 PM · Jul 9, 2026 · 6.19M Views

183 Replies · 2.67K Reposts · 60.3K Likes

Best,
Manas