{"schema_version":1,"generated_at":"2026-11-20T02:29:47.174778+00:00","article_count":32,"articles":[{"id":"journals:10.1039/d6sc06681j","kind":"journals","source":"Chemical Science","title":"Automated multimodal mining of structured data locked in the figures and tables of flow chemistry literature","url":"https://doi.org/10.1039/d6sc06681j","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.1039%2Fd6sc06681j","date":"2026-11-19T17:24:42Z","timestamp":1795109082,"categories":[],"topic_ids":[],"keywords":[],"matched_keywords":["machine learning","molecular"],"matched_tags":["machine_learning","generic_chem"],"doi":"10.1039/d6sc06681j","external_id":null,"pdf_url":null,"code_url":null,"code_host":null,"authors":["Wenyuan Zhao","Xianzhu Zhong","Xavier Jouffroy","Yosuke Ashikari","Hiromichi V. Miyagishi","Mohmmad S. Qenawy","Simeng Wang","Yirui Peng","Mikail Raffano Munsyi","Aiichiro Nagaki"],"journal":"Chemical Science","publisher":"RSC","impact_factor":7.5,"abstract":"Flow chemistry literature locks much of its quantitative process data inside figures and tables that show molecular structures as images, including the low-yield and failed reactions that machine learning most needs.","source_metadata":{"short_name":"Chem. Sci.","source":"crossref","abstract_source":null,"abstract_quality":null}},{"id":"journals:10.1039/d6sc06542b","kind":"journals","source":"Chemical Science","title":"Generative models in industrial drug discovery: perspectives from REINVENT","url":"https://doi.org/10.1039/d6sc06542b","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.1039%2Fd6sc06542b","date":"2026-11-19T15:02:58Z","timestamp":1795100578,"categories":[],"topic_ids":[],"keywords":["Generative models","REINVENT"],"matched_keywords":["Generative models","REINVENT","drug","chemical"],"matched_tags":["generative_model","reinforcement_learning","generic_chem"],"doi":"10.1039/d6sc06542b","external_id":null,"pdf_url":null,"code_url":null,"code_host":null,"authors":["Jon Paul Janet"],"journal":"Chemical Science","publisher":"RSC","impact_factor":7.5,"abstract":"Generative models have become established tools for exploring chemical space in early-stage drug discovery and are among the most important practical applications of AI for science. Numerous studies have reported...","source_metadata":{"short_name":"Chem. Sci.","source":"crossref","abstract_source":null,"abstract_quality":null}},{"id":"feeds:https://macinchem.org/2026/11/19/chemistry-careers/","kind":"feeds","source":"Macs in Chemistry (Macinchem Blog)","title":"Chemistry careers","url":"https://macinchem.org/2026/11/19/chemistry-careers/","detail_url":"/chemradar/article?u=https%3A%2F%2Fmacinchem.org%2F2026%2F11%2F19%2Fchemistry-careers%2F","date":"2026-11-19T12:07:18+00:00","timestamp":1795090038,"categories":["Blog"],"topic_ids":[],"keywords":[],"matched_keywords":[],"matched_tags":[],"doi":null,"external_id":null,"pdf_url":null,"code_url":null,"code_host":null,"authors":[],"journal":null,"publisher":null,"impact_factor":null,"abstract":"","source_metadata":{"source":"Macs in Chemistry (Macinchem Blog)","published_utc":"2026-11-19T12:07:18+00:00","seen_at":"2026-11-19T14:20:39.947520+00:00"}},{"id":"preprints:2611.25452v1","kind":"preprints","source":"arXiv","title":"FamilySpace: From Emergent Chemical Semantics to Controllable Crystal Generation","url":"https://arxiv.org/abs/2611.25452v1","detail_url":"/chemradar/article?u=https%3A%2F%2Farxiv.org%2Fabs%2F2611.25452v1","date":"2026-11-19T03:32:41Z","timestamp":1795059161,"categories":["Design de novo"],"topic_ids":["denovo"],"keywords":["generative models","reinforcement learning","variational autoencoder","VAE"],"matched_keywords":["generative models","reinforcement learning","variational autoencoder","VAE","Chemical"],"matched_tags":["generative_model","reinforcement_learning","vae","generic_chem"],"doi":null,"external_id":"2611.25452v1","pdf_url":"https://arxiv.org/pdf/2611.25452v1","code_url":"https://github.com/hhw11/FamilySpace","code_host":"GitHub","authors":["Haowei Hua","Irwin King","Wanyu Lin"],"journal":null,"publisher":null,"impact_factor":null,"abstract":"Latent crystal generative models typically combine a variational autoencoder (VAE) for learning atom-wise representations with a flow model that transports a simple source distribution to the learned latent distribution. However, previous studies have mainly focused on the quality of generated crystals, leaving the structure and generative utility of the atom-wise latent space largely unexplored. In this work, we show that even a vanilla crystal VAE implicitly organizes atom-wise latent representations according to elemental families, with different families occupying largely separated regions of the latent space. We term this phenomenon emergent chemical semantics. This structure suggests that sampling from a family-specific latent region can bias the decoder toward atoms from the corresponding elemental family. Building on this insight, we propose FamilySpace, a framework for controlling the elemental-family composition of generated crystals. FamilySpace assigns each elemental family a distinct source component, forming a family-indexed multi-source distribution, and trains a flow model using source-target pairs matched by family. In inference, a user-specified family template selects the source component for each atom-wise representation, enabling direct control over the family composition of generated crystals. Experiments on MP-20 and MPTS-52 show that FamilySpace achieves competitive crystal-generation performance while matching prescribed elemental-family compositions. Moreover, selected family templates increase the fraction of generated crystals predicted to be stable or metastable, demonstrating that latent-space composition control can support targeted exploration of promising chemical regions without additional optimization techniques such as reinforcement learning. Code is available at https://github.com/hhw11/FamilySpace","source_metadata":{"categories":["cs.LG","cond-mat.mtrl-sci"],"code_url":"https://github.com/hhw11/FamilySpace","code_status":"found"}},{"id":"journals:10.1021/acsomega.6c07978","kind":"journals","source":"ACS Omega","title":"3,4-Bis(benzyloxy)benzaldehyde-Derived\nHydrazones:\nSynthesis, α-Amylase-Selective Inhibition, Antioxidant\nProfiling, and In Silico Evaluation","url":"https://doi.org/10.1021/acsomega.6c07978","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.1021%2Facsomega.6c07978","date":"2026-11-19T00:00:00+00:00","timestamp":1795046400,"categories":["Docking & Screening"],"topic_ids":["docking"],"keywords":["Molecular docking","IC50"],"matched_keywords":["Molecular docking","compounds","compound","drug","Molecular","IC50"],"matched_tags":["docking","generic_chem","bioactivity"],"doi":"10.1021/acsomega.6c07978","external_id":null,"pdf_url":null,"code_url":null,"code_host":null,"authors":["Hanifa Fatullayev","Ismail Celik","Leyla Paşayeva"],"journal":"ACS Omega","publisher":"ACS","impact_factor":null,"abstract":"Carbohydrate-hydrolyzing enzymes (α-amylase and α-glucosidase) are clinically validated targets for controlling postprandial hyperglycemia in type 2 diabetes mellitus (T2DM), and oxidative stress is a recognized contributor to disease progression. Herein, a series of hydrazone Schiff bases derived from 3,4-bis(benzyloxy)benzaldehyde were synthesized by acid-catalyzed condensation of the aldehyde with para-substituted phenylhydrazines and were characterized by 1H and 13C NMR spectroscopy, FT-IR spectroscopy, and high-resolution mass spectrometry; chromatographic purity was established by RP-HPLC (94.5–98.6 area % at 350 nm). The series was evaluated for in vitro inhibition of porcine pancreatic α-amylase and yeast α-glucosidase and for antioxidant capacity using four complementary assays (DPPH, ABTS, FRAP, and CUPRAC). The compounds displayed a pronounced α-amylase preference: compound 9 (IC50 = 160.40 ± 2.05 μg/mL) and compound 7 (IC50 = 166.01 ± 2.64 μg/mL) were the most potent and exceeded acarbose under the assay conditions (IC50 = 232.66 ± 4.93 μg/mL); on a molar basis, both are essentially equipotent with the reference drug (345.2 ± 4.4 and 378.6 ± 6.0 μM versus 360.4 ± 7.6 μM), and IC50 values are reported in both units throughout (Table 1). α-Glucosidase inhibition was moderate, with the para-chloro analogue 3 being the most active (IC50 = 83.13 ± 6.53 μg/mL; 187.7 ± 14.7 μM). The selectivity index (IC50 α-glucosidase/IC50 α-amylase) spans 0.22 for 3 to 2.38 for 9, so a single para-substituent determines which carbohydrase is preferred. Antioxidant behavior was assay-dependent: compounds 1 and 4 delivered the highest ferric-reducing power (FRAP = 1854.0 ± 1.5 and 1786.7 ± 5.2 μmol Fe2+/g), and compound 5 gave the strongest CUPRAC response (280.95 ± 3.43 mg GAE/g). Molecular docking against α-amylase (PDB 1B2Y) and α-glucosidase (PDB 3TOP), together with SwissADME, QikProp, and ADMETlab profiling, provides a structural rationale for the observed substituent effects. These results define a clear, substituent-driven selectivity relationship within a lipophilic dibenzyloxy hydrazone framework and establish the DBZO-hydrazone scaffold as a tractable system for further optimization.","source_metadata":{"short_name":"ACS Omega","source":"crossref","abstract_source":null,"abstract_quality":null}},{"id":"journals:10.1021/acs.jcim.6c03192","kind":"journals","source":"Journal of Chemical Information and Modeling","title":"Comment on “Hybrid\nGraph–Machine Learning\nFramework for Accurate and Interpretable Band Gap Prediction”","url":"https://doi.org/10.1021/acs.jcim.6c03192","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.1021%2Facs.jcim.6c03192","date":"2026-11-19T00:00:00+00:00","timestamp":1795046400,"categories":["Property Prediction"],"topic_ids":["property"],"keywords":["graph neural networks","GNN","DFT"],"matched_keywords":["neural networks","graph neural networks","GNN","Machine Learning","DFT"],"matched_tags":["neural_network","gnn","machine_learning","dft"],"doi":"10.1021/acs.jcim.6c03192","external_id":null,"pdf_url":null,"code_url":null,"code_host":null,"authors":["Shambhu Bhandari Sharma"],"journal":"Journal of Chemical Information and Modeling","publisher":"ACS","impact_factor":5.3,"abstract":"Aydin et al. reported R2 = 0.921 for band gap prediction using hybrid graph neural networks. Reproduction shows this performance is inflated by data leakage from post-DFT proxies (is_metal, is_gap_direct) and embedding contamination. Controlled ablations reveal a genuine GNN gain of 0.08–0.11 in R2, while leakage contributes ∼0.06, lowering honest performance to 0.82–0.86 R2 (MAE 0.28–0.30 eV).","source_metadata":{"short_name":"JCIM","source":"crossref","abstract_source":null,"abstract_quality":null}},{"id":"preprints:10.26434/chemrxiv.15002852/v5","kind":"preprints","source":"ChemRxiv","title":"De Novo Design of Enantioselective Artificial Metalloenzyme Photocatalysts","url":"https://doi.org/10.26434/chemrxiv.15002852/v5","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.26434%2Fchemrxiv.15002852%2Fv5","date":"2026-11-19T00:00:00Z","timestamp":1795046400,"categories":[],"topic_ids":[],"keywords":["De Novo Design"],"matched_keywords":["De Novo Design"],"matched_tags":["de_novo_design"],"doi":"10.26434/chemrxiv.15002852/v5","external_id":"10.26434/chemrxiv.15002852/v5","pdf_url":null,"code_url":null,"code_host":null,"authors":["Sandip Mishra","Payal","Declan Evans","Kingsley Bortey","Husayn Bootwala","Giovanni Gonzalez-Gutierrez","Ricardo Javier Vázquez","David Baker","Jared C Lewis"],"journal":null,"publisher":null,"impact_factor":null,"abstract":"","source_metadata":{"date_field":"posted","keywords":[]}},{"id":"journals:10.1021/acs.jcim.6c02673","kind":"journals","source":"Journal of Chemical Information and Modeling","title":"DeepChem-DEL: An\nOpen-Source Framework for Reproducible\nDEL Modeling and Benchmarking","url":"https://doi.org/10.1021/acs.jcim.6c02673","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.1021%2Facs.jcim.6c02673","date":"2026-11-19T00:00:00+00:00","timestamp":1795046400,"categories":["Cheminformatics","Docking & Screening","Design de novo","Library Design"],"topic_ids":["cheminformatics","docking","denovo","libraries"],"keywords":["ChemBERTa","random forest","denoising","DNA encoded libraries"],"matched_keywords":["ChemBERTa","random forest","Machine learning","ML","denoising","DNA-encoded libraries","chemical","compounds"],"matched_tags":["pretrained_chem_models","ensemble","machine_learning","diffusion","dna_encoded_library","generic_chem"],"doi":"10.1021/acs.jcim.6c02673","external_id":null,"pdf_url":null,"code_url":null,"code_host":null,"authors":["Aryan Amit Barsainyan","Riya Singh","Abhiraj Pravin Mengade","Rida Irfan","Bharath Ramsundar"],"journal":"Journal of Chemical Information and Modeling","publisher":"ACS","impact_factor":5.3,"abstract":"DNA-encoded libraries (DELs) allow screening of extremely large chemical spaces by tagging compounds with DNA barcodes. Machine learning (ML) techniques have enabled the extraction of meaningful chemical patterns from noisy, target-specific DEL binding data. However, existing tools for DEL machine learning (DEL-ML) remain fragmented, which limits systematic construction and reproducibility of modeling pipelines. Moreover, this fragmentation hinders integration of new models, such as chemical foundation models, into DEL workflows alongside classical approaches. To address these challenges in DEL modeling, we introduce DeepChem-DEL, an open-source suite of reproducible ML workflows for DEL modeling built on the DeepChem ecosystem. DeepChem-DEL integrates (i) a configurable denoising pipeline with (ii) modular DeepChem workflows for enrichment prediction, hit identification, and benchmarking across diverse model classes, including chemical foundation models enabled through extensions to the open-source ChemBERTa-3 framework. We evaluate DeepChem-DEL on the KinDEL and BELKA data sets, reproducing key baselines and enabling direct comparison across model classes. Our experiments demonstrate that leveraging chemical foundation models yields performance competitive with, and in several cases exceeding, previously explored approaches such as graph convolutional networks and random forest models. Together, these results demonstrate that DeepChem-DEL is a scalable and extensible platform for DEL-ML, while highlighting the potential of chemical foundation models for improved DEL modeling and prioritization.","source_metadata":{"short_name":"JCIM","source":"crossref","abstract_source":null,"abstract_quality":null}},{"id":"preprints:10.26434/chemrxiv.15011551/v2","kind":"preprints","source":"ChemRxiv","title":"Design and Benchmarking of CK2α′-Preferring Inhibitors for Huntington’s Disease: The Limits of Structure-Based Isoform Selectivity","url":"https://doi.org/10.26434/chemrxiv.15011551/v2","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.26434%2Fchemrxiv.15011551%2Fv2","date":"2026-11-19T00:00:00Z","timestamp":1795046400,"categories":[],"topic_ids":[],"keywords":["ChEMBL","kinase"],"matched_keywords":["ChEMBL","compounds","Molecular","Protein","kinase"],"matched_tags":["chem_benchmarks","generic_chem","biology"],"doi":"10.26434/chemrxiv.15011551/v2","external_id":"10.26434/chemrxiv.15011551/v2","pdf_url":null,"code_url":null,"code_host":null,"authors":["Tarun Tarun"],"journal":null,"publisher":null,"impact_factor":null,"abstract":"Protein kinase CK2 acts through two catalytic isoforms, CK2α (A1) and CK2α′ (A2), whose ATP-binding sites differ at only two hinge positions, making isoform-selective inhibition difficult. We tested whether a standard structure-based pipeline can design and rank CK2α′-preferring compounds. In a paired benchmark of 210 compounds measured against both isoforms, the inter-isoform docking score was unrelated to the experimental selectivity (Spearman ρ = −0.22), and even within the benzo[c][2,6]naphthyridine scaffold, the scoring reversed the order set by the ChEMBL activities. The only reliable determinant of preference, a two-carbon linker between the anilino nitrogen and the pendant ring of this scaffold, emerged from the measured activities rather than from any computed score. Molecular-dynamics ensembles of the two isoforms overlapped (0.88 Å between versus 0.72–0.96 Å within), placing the selectivity signal below the resolution of the methods. From an enumerated series, EV043 showed the largest end-point free-energy preference for CK2α′ (ΔΔG b -7.74 kcal/mol), although this value remained sensitive to the starting pose in CK2α, while the non-selective inhibitor CX-4945 was reproduced correctly (+0.08 kcal/mol). Distinguishing CK2α′ from CK2α therefore requires alchemical free-energy methods supported by experimental validation.","source_metadata":{"date_field":"posted","keywords":[]}},{"id":"journals:10.1021/acs.jafc.6c11247","kind":"journals","source":"Journal of Agricultural and Food Chemistry","title":"Design, Synthesis,\nand Antifungal Evaluation of Novel\nPyridine-3-amide Derivatives Containing Phenylthiazole Moiety as Potential\nSuccinate Dehydrogenase Inhibitors","url":"https://doi.org/10.1021/acs.jafc.6c11247","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.1021%2Facs.jafc.6c11247","date":"2026-11-19T00:00:00+00:00","timestamp":1795046400,"categories":["Docking & Screening"],"topic_ids":["docking"],"keywords":["Molecular docking","EC50"],"matched_keywords":["Molecular docking","compound","Molecular","EC50"],"matched_tags":["docking","generic_chem","bioactivity"],"doi":"10.1021/acs.jafc.6c11247","external_id":null,"pdf_url":null,"code_url":null,"code_host":null,"authors":["Jin Wang","Zhiyuan Xu","Sen Huang","Huochun Ye","Fadi Zhu","Jing Zhang","Gang Feng"],"journal":"Journal of Agricultural and Food Chemistry","publisher":"ACS","impact_factor":null,"abstract":"In search of novel succinate dehydrogenase inhibitor fungicides (SDHIs) for the management of plant fungal diseases, 29 novel pyridine-3-amide derivatives containing a phenylthiazole moiety were designed and synthesized. Among them, compound H18 showed the strongest antifungal activity against B. cinerea, with an EC50 value of 0.06 μg/mL, substantially lower than that of boscalid (1.79 μg/mL). Spearman’s correlation analysis revealed no cross-resistance between H18 and boscalid (ρ = −0.1743, P = 0.6082). Moreover, H18 exhibited strong protective and curative efficacy of 79.98% and 75.57%, respectively, against B. cinerea on tomatoes at a concentration of 100 μg/mL, superior to that of boscalid. Mechanistic studies demonstrated that H18 inhibited SDH activity, leading to ROS accumulation, disruption of mitochondrial energy metabolism, and fungal cell death. Molecular docking suggested that the flexible conformations of the phenylthiazole moiety in H18 may contribute to its potential resistance-overcoming capacity. These findings identify H18 as a promising new SDHI candidate for B. cinerea control, warranting further investigation.","source_metadata":{"short_name":"J. Agric. Food Chem.","source":"crossref","abstract_source":null,"abstract_quality":null}},{"id":"journals:10.1021/acs.chemrev.6c00374","kind":"journals","source":"Chemical Reviews","title":"Development\nand Application of Nanopore Direct RNA\nSequencing","url":"https://doi.org/10.1021/acs.chemrev.6c00374","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.1021%2Facs.chemrev.6c00374","date":"2026-11-19T00:00:00+00:00","timestamp":1795046400,"categories":[],"topic_ids":[],"keywords":[],"matched_keywords":["machine learning","protein"],"matched_tags":["machine_learning","biology"],"doi":"10.1021/acs.chemrev.6c00374","external_id":null,"pdf_url":null,"code_url":null,"code_host":null,"authors":["Miten Jain","Robin Abu-Shumays","Hugh E. Olsen","Mark Akeson"],"journal":"Chemical Reviews","publisher":"ACS","impact_factor":55.8,"abstract":"Individual RNA strands can be sequenced directly using nanoscale protein pores. This process, Nanopore Direct RNA Sequencing (Nanopore DRS), uses a helicase motor protein to regulate RNA displacement through the nanopore orifice in single nucleotide steps. Partial blockade of the monovalent ionic current is recorded at 4 kHz and converted to a string of RNA nucleotide basecalls using machine learning models. In this review, we discuss the development of Nanopore DRS from 1996 to the present. This includes performance metrics, evolution of machine learning models, and applications of Nanopore DRS to major RNA classes across kingdoms. Advantages of the technology are highlighted with special emphasis on nucleotide modifications along intact single strands. Shortcomings of the technology are also considered along with possible solutions. We close with forward looking Nanopore DRS applications: Single cell RNA sequencing; remote monitoring of infectious agents; principled understanding of nanopore ionic current patterns using all-atoms simulations; global food security; and clinical applications.","source_metadata":{"short_name":"Chem. Rev.","source":"crossref","abstract_source":null,"abstract_quality":null}},{"id":"journals:10.1038/s41598-026-81415-y","kind":"journals","source":"Scientific Reports","title":"Evaluation of sensing response of H2S sensing materials toward H2S molecules using artificial neural network and response surface methodology","url":"https://doi.org/10.1038/s41598-026-81415-y","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.1038%2Fs41598-026-81415-y","date":"2026-11-19T00:00:00+00:00","timestamp":1795046400,"categories":[],"topic_ids":[],"keywords":[],"matched_keywords":["neural network","molecules"],"matched_tags":["neural_network","generic_chem"],"doi":"10.1038/s41598-026-81415-y","external_id":null,"pdf_url":null,"code_url":null,"code_host":null,"authors":["Fatemeh Hamidipour","Hadiseh Masoumi","Ahmad Rahbar-Kelishami"],"journal":"Scientific Reports","publisher":"Nature","impact_factor":null,"abstract":"","source_metadata":{"short_name":"Sci. Rep.","source":"crossref","abstract_source":null,"abstract_quality":null}},{"id":"preprints:10.26434/chemrxiv.15012462/v1","kind":"preprints","source":"ChemRxiv","title":"Exploring Boltzmann Distributions via Kernel-Based Fokker–Planck Ensemble Transport","url":"https://doi.org/10.26434/chemrxiv.15012462/v1","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.26434%2Fchemrxiv.15012462%2Fv1","date":"2026-11-19T00:00:00Z","timestamp":1795046400,"categories":[],"topic_ids":[],"keywords":["Transformer","free energy"],"matched_keywords":["Transformer","free energy"],"matched_tags":["transformer","thermodynamics"],"doi":"10.26434/chemrxiv.15012462/v1","external_id":"10.26434/chemrxiv.15012462/v1","pdf_url":null,"code_url":null,"code_host":null,"authors":["Xiaoxuan Yu","Zichen Sun","Jun Zhang","Yi Qin Gao"],"journal":null,"publisher":null,"impact_factor":null,"abstract":"Local trajectory-based samplers can relax within a visited basin long before finite trajectories cross barriers to other Boltzmann-relevant regions. We introduce the Fokker–Planck Transformer (FPT), a training-free interacting-ensemble method that makes the instantaneous ensemble distribution a dynamical variable, shifting exploration from trajectory-wise barrier crossing toward distribution-level redistribution. FPT discretizes the current-velocity representation of overdamped Fokker–Planck evolution by combining energetic drift with a kernel estimate of the ensemble score. The continuum flow has the Boltzmann density as its stationary target and monotonically dissipates configurational free energy and relative entropy. A Gaussian row-mass selector couples increasing local occupancy to a vanishing spatial scale without a separate bandwidth limit. For its fresh explicit-Euler realization, pathwise score residuals propagate to distributional and observable errors, while a weak-flux argument gives conditional fixed-horizon convergence to the Fokker–Planck solution. For this realization, a controlled two-dimensional calculation tracks transport from a localized Gaussian toward equilibrium against a numerically propagated Fokker–Planck reference. In a three-dimensional periodic landscape, an exact Fokker–Planck reference shows that a shared-bandwidth approximation populates distant wells far ahead of the continuum solution, which accounts for its faster support discovery; the transported ensemble runs warm, and a Hastings-corrected continuation from its endpoint restores the energy distribution within two steps. Further tests in a 100-dimensional landscape, alanine decapeptide, and 190 docking systems show rapid expansion of Boltzmann-relevant support, with the clearest gains over matched local controls when barriers or configurational complexity limit global exploration. Overall, FPT provides thermodynamically grounded finite-budget support discovery and can seed target-preserving sampling.","source_metadata":{"date_field":"posted","keywords":[]}},{"id":"journals:10.1038/s43588-026-01082-x","kind":"journals","source":"Nature Computational Science","title":"Generative molecular design needs oracle grounding","url":"https://doi.org/10.1038/s43588-026-01082-x","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.1038%2Fs43588-026-01082-x","date":"2026-11-19T00:00:00+00:00","timestamp":1795046400,"categories":["Design de novo"],"topic_ids":["denovo"],"keywords":["Generative molecular design"],"matched_keywords":["Generative molecular design","molecular"],"matched_tags":["generative_drug_design","generic_chem"],"doi":"10.1038/s43588-026-01082-x","external_id":null,"pdf_url":null,"code_url":null,"code_host":null,"authors":["Sarina Kopf","Philippe Schwaller"],"journal":"Nature Computational Science","publisher":"Nature","impact_factor":18.3,"abstract":"","source_metadata":{"short_name":"Nat. Comput. Sci.","source":"crossref","abstract_source":null,"abstract_quality":null}},{"id":"journals:10.1021/acsomega.6c10360","kind":"journals","source":"ACS Omega","title":"In Silico\n                    Identification of Peptide\nPharmacophoric Determinants and Complementary Binding Hotspots on\nFungal Grr1","url":"https://doi.org/10.1021/acsomega.6c10360","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.1021%2Facsomega.6c10360","date":"2026-11-19T00:00:00+00:00","timestamp":1795046400,"categories":["Docking & Screening","Free Energy & MD"],"topic_ids":["docking","freeenergy"],"keywords":["pharmacophore modeling","pharmacophore mapping","binding free energy","molecular features","molecular dynamics"],"matched_keywords":["pharmacophore modeling","pharmacophore mapping","binding free energy","molecular features","molecular dynamics","molecular","ligand","ligands","protein"],"matched_tags":["pharmacophore","thermodynamics","molecular_descriptor","molecular_dynamics","generic_chem","biology"],"doi":"10.1021/acsomega.6c10360","external_id":null,"pdf_url":null,"code_url":null,"code_host":null,"authors":["Mateus Dias-Oliveira","Henrique Bridi","Rogerio Margis","Geancarlo Zanatta"],"journal":"ACS Omega","publisher":"ACS","impact_factor":null,"abstract":"The growing burden of human and plant fungal diseases, compounded by limited therapeutic options and emerging antifungal resistance, underscores the need to identify novel molecular targets for antifungal intervention. In this context, the fungal F-box protein Grr1 acts as the substrate-recognition component of the SCF ubiquitin ligase complex, binding phosphorylated degrons to direct substrate degradation and regulate key cellular processes. However, the structural and energetic determinants governing phosphodegron recognition by the Grr1 leucine-rich repeat (LRR) domain remain poorly characterized. Here, we investigated the molecular basis of Grr1 recognition of phosphodegrons derived from the substrates Cln2 and Gic2 and evaluated whether phosphomimetic substitutions could preserve their interaction with the LRR domain. Structural modeling, molecular dynamics simulations, binding free-energy calculations, per-residue energy decomposition, and pharmacophore modeling were integrated to characterize the Grr1-degron interfaces and identify their major recognition determinants. Phosphorylation was required to maintain stable association of both degrons with the concave surface of the Grr1 LRR domain. Per-residue energy analysis revealed a conserved basic recognition surface comprising Arg418, Arg496, Lys498, Arg550, Lys608, Arg680, and Arg709, with Lys498, Arg550, Arg680, and Arg709 emerging as the principal energetic contributors shared by both complexes. Together, these residues form an electropositive environment that coordinates the degron phosphate groups, while pharmacophore mapping identified complementary molecular features associated with ligand recognition. Remarkably, substitution of the phosphorylated residues with aspartate preserved the calculated binding free energy of both degrons, whereas glutamate substitution disrupted stable binding, indicating that phosphomimetic recognition depends not merely on negative charge but critically on its precise spatial positioning within the recognition interface. Collectively, these findings define a conserved structural and energetic mechanism of phosphodegron recognition by fungal Grr1 and identify molecular features that can be exploited for ligand development. This framework supports the rational design of phosphomimetic peptides and other competitive ligands targeting the Grr1 substrate-recognition interface and provides a foundation for exploring this conserved regulatory system as a potential strategy for antifungal intervention.","source_metadata":{"short_name":"ACS Omega","source":"crossref","abstract_source":null,"abstract_quality":null}},{"id":"journals:10.1038/s41598-026-78564-5","kind":"journals","source":"Scientific Reports","title":"In silico screening of FDA-approved drugs for elimination of senescent cells via senescence-associated anti-apoptotic pathway inhibition","url":"https://doi.org/10.1038/s41598-026-78564-5","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.1038%2Fs41598-026-78564-5","date":"2026-11-19T00:00:00+00:00","timestamp":1795046400,"categories":[],"topic_ids":[],"keywords":["In silico screening"],"matched_keywords":["In silico screening","drugs"],"matched_tags":["virtual_screening","generic_chem"],"doi":"10.1038/s41598-026-78564-5","external_id":null,"pdf_url":null,"code_url":null,"code_host":null,"authors":["Emad Manni"],"journal":"Scientific Reports","publisher":"Nature","impact_factor":null,"abstract":"","source_metadata":{"short_name":"Sci. Rep.","source":"crossref","abstract_source":null,"abstract_quality":null}},{"id":"journals:10.1021/acs.jafc.6c10849","kind":"journals","source":"Journal of Agricultural and Food Chemistry","title":"Integrative Omics\nReveal the Antiviral Mechanism of\nGossypol and Guide the Construction of a Chitooligosaccharide–Gossypol\nNanocomposite","url":"https://doi.org/10.1021/acs.jafc.6c10849","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.1021%2Facs.jafc.6c10849","date":"2026-11-19T00:00:00+00:00","timestamp":1795046400,"categories":["Docking & Screening"],"topic_ids":["docking"],"keywords":["Molecular docking","molecular dynamics"],"matched_keywords":["Molecular docking","Molecular","molecular dynamics"],"matched_tags":["docking","generic_chem","molecular_dynamics"],"doi":"10.1021/acs.jafc.6c10849","external_id":null,"pdf_url":null,"code_url":null,"code_host":null,"authors":["Wenkang Li","Weifeng Wu","Huimin Chang","Shuo Qin","Maozhi Ren","Peng Wang","Yan Zhao","Jun Li","Fuguang Li","Wenjing Wang"],"journal":"Journal of Agricultural and Food Chemistry","publisher":"ACS","impact_factor":null,"abstract":"Gossypol (GOS), a unique cotton sesquiterpenoid, exhibits broad antiviral activity against animal viruses, but its specific effects on plant viruses remain largely unexplored. Using tobacco mosaic virus (TMV) and cucumber mosaic virus (CMV) models, we demonstrated that GOS treatment significantly reduces viral RNA accumulation and alleviates systemic symptoms. Integrating multiple omics approaches, findings suggest that GOS rebalances host hormones by repressing biosynthesis and activating catabolism, curbing excess salicylic acid (SA) and abscisic acid (ABA) while restoring indole acetic acid (IAA) to optimize resource allocation between growth and defense. Additionally, the transcriptional upregulation of antiviral sesquiterpene genes closely aligns with the targeted metabolite accumulation. Molecular docking, molecular dynamics, and surface plasmon resonance (SPR) assays reveal that GOS directly targets viral RNA-dependent RNA polymerase (RdRp). Collectively, GOS mitigates plant viral infections through polymerase targeting and defense reprogramming. These insights inspired a highly efficient GOS–chitooligosaccharide (COS) nanocomplex for crop protection in sustainable agriculture.","source_metadata":{"short_name":"J. Agric. Food Chem.","source":"crossref","abstract_source":null,"abstract_quality":null}},{"id":"preprints:10.26434/chemrxiv.15012474/v1","kind":"preprints","source":"ChemRxiv","title":"Linear-Algebraic Audit of Biomedical Matrices: Five Empirical Case Studies in Regularized Inference","url":"https://doi.org/10.26434/chemrxiv.15012474/v1","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.26434%2Fchemrxiv.15012474%2Fv1","date":"2026-11-19T00:00:00Z","timestamp":1795046400,"categories":[],"topic_ids":[],"keywords":["chemometric"],"matched_keywords":["chemometric"],"matched_tags":["chemometrics"],"doi":"10.26434/chemrxiv.15012474/v1","external_id":"10.26434/chemrxiv.15012474/v1","pdf_url":null,"code_url":null,"code_host":null,"authors":["Petr Avtonomov"],"journal":null,"publisher":null,"impact_factor":null,"abstract":"High-throughput biomedical assays—ranging from gene expression microarrays to mass spectrometry and clinical registries—routinely exhibit severe rank deficiency (N ≪ P ), zero-inflation, and extreme sample outliers. Standard chemometric and bioinformatics pipelines apply variance auto-scaling, simplex normalization, and bilinear projections (PCA/PLS-DA) without verifying operator conditioning or modal localization. Grounded in inverse problem theory, this empirical study subjects five public biomedical datasets to a sequential four-step linear-algebraic audit evaluating residual Gram conditioning (κ), Effective Degrees of Freedom (DoF), singular spectra (Σ), and Participation Ratios (PR). Across five empirical regimes—Golub leukemia microarrays (72×7129), a large-scale oncology cohort (1545×10935), zero-inflated NCI Arcene mass spectra (200 × 10000), Pomeroy central nervous system tumor profiles (60 × 7129), and the Cleveland cardiology registry (270 × 13)—we demonstrate how uncalibrated preprocessing inflates background noise threefold, collapses leading components onto fewer than 7 outlier samples, and bends multi-batch artifacts into parabolic horseshoes. Removing just 0.6 percent of leverage outliers identified by PR restores sample coherence nearly ninety-fold (PR : 6.4 → 548.1), providing a rigorous diagnostic standard for biomedical data hygiene.","source_metadata":{"date_field":"posted","keywords":[]}},{"id":"journals:10.1038/s41598-026-80158-0","kind":"journals","source":"Scientific Reports","title":"Machine learning and molecular validation reveal the PI3K/AKT/mTOR inflammatory axis for therapeutic targeting in severe COVID-19","url":"https://doi.org/10.1038/s41598-026-80158-0","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.1038%2Fs41598-026-80158-0","date":"2026-11-19T00:00:00+00:00","timestamp":1795046400,"categories":[],"topic_ids":[],"keywords":[],"matched_keywords":["Machine learning","molecular"],"matched_tags":["machine_learning","generic_chem"],"doi":"10.1038/s41598-026-80158-0","external_id":null,"pdf_url":null,"code_url":null,"code_host":null,"authors":["Sahar Samsami","Hiva Alipanah","Alireza Faghani Ghodrat","Mohammad Mehdi Naghizadeh","Seyedeh-Samira Ashrafmansouri","Ali Ghanbariasad"],"journal":"Scientific Reports","publisher":"Nature","impact_factor":null,"abstract":"","source_metadata":{"short_name":"Sci. Rep.","source":"crossref","abstract_source":null,"abstract_quality":null}},{"id":"journals:10.1038/s41598-026-77567-6","kind":"journals","source":"Scientific Reports","title":"Rapid quantification of patchouli alcohol in patchouli oil (Pogostemon Cablin) using near infrared spectroscopy with multivariate and machine learning approaches","url":"https://doi.org/10.1038/s41598-026-77567-6","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.1038%2Fs41598-026-77567-6","date":"2026-11-19T00:00:00+00:00","timestamp":1795046400,"categories":["Spectra & Analytical"],"topic_ids":["chemometrics"],"keywords":[],"matched_keywords":["infrared spectroscopy with multivariate and machine learning","machine learning"],"matched_tags":["analytical_chemistry_ai","machine_learning"],"doi":"10.1038/s41598-026-77567-6","external_id":null,"pdf_url":null,"code_url":null,"code_host":null,"authors":["Nurul Ainina","Sarifah Nurjanah","Uma Rani Sinniah","Rahmat Hidayat","Euis Julaeha","Muhammad Achirul Nanda","Mimin Muhaemin"],"journal":"Scientific Reports","publisher":"Nature","impact_factor":null,"abstract":"Patchouli oil ( Pogostemon Cablin ) is an essential oil primarily used as a fixative ingredient in the perfumery and cosmetic industries. Its major component, patchouli alcohol (PA), is recognized as a key quality parameter, with higher contents being desirable. However, the conventional method for quantifying PA, namely gas chromatography-mass spectrometry (GC-MS), is relatively expensive, time-consuming, and destructive. Therefore, this study aimed to explore near-infrared (NIR) spectroscopy combined with multivariate and machine learning algorithms, including partial least squares regression (PLSR), support vector regression (SVR), and a hybrid method, partial least squares-support vector regression (PLS-SVR). The best-performing models was obtained using PLSR combined with Savitzky-Golay 2nd derivative (SG 2nd) preprocessing in a single-split evaluation, achieving R 2 CV = 0.9094, RMSECV = 2.3471, R 2 P = 0.9251, RMSEP = 2.3099, and RPD = 3.676. The robustness of this model was further confirmed using Monte-Carlo double cross-validation (MCCV-DCV), yielding a mean of R 2 P and RMSEP at 0.8909 ± 0.0417 and 2.4329 ± 0.4402, respectively, across thirty repeated iterations. A paired statistical comparison between PLSR and PLS-SVR further showed no significant difference in RMSEP (Wilcoxon singed-rank test, p-value = 0.1347), while SVR performed significantly worse ( p < 0.05). In conclusion, NIR spectroscopy using PLSR method, could be used as a rapid, simple, low-cost alternative method for estimating PA content in patchouli oil.","source_metadata":{"short_name":"Sci. Rep.","source":"crossref","abstract_source":null,"abstract_quality":null}},{"id":"preprints:10.26434/chemrxiv.15012472/v1","kind":"preprints","source":"ChemRxiv","title":"Self-Consistent Induced Dipoles for MACE-POLAR in Molecular Embedding and Classical Dynamics","url":"https://doi.org/10.26434/chemrxiv.15012472/v1","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.26434%2Fchemrxiv.15012472%2Fv1","date":"2026-11-19T00:00:00Z","timestamp":1795046400,"categories":["ML Potentials"],"topic_ids":["potentials"],"keywords":["MACE","DFT"],"matched_keywords":["MACE","DFT","Molecular"],"matched_tags":["mlip_tools","dft","generic_chem"],"doi":"10.26434/chemrxiv.15012472/v1","external_id":"10.26434/chemrxiv.15012472/v1","pdf_url":null,"code_url":null,"code_host":null,"authors":["Jianxin Cheng"],"journal":null,"publisher":null,"impact_factor":null,"abstract":"We couple the native Gaussian multipoles of MACE-POLAR-1-M to classical induced dipoles and compare the resulting solute densities, fields and forces with quantum references. The density-driven fixed point has a nonzero energy derivative with respect to its dipoles. An adjoint recovers forces on the converged surface to within 2.13 × 10−7 eV/Å of reconverged finite differences in twelve force-test configurations. A separate density comparison treats both partners quantum mechanically at twelve geometries of six small complexes. At a common Gaussian width of 1.5 Å, enabling mutual induction changes the native MACE solute density by at most 6.81 × 10−4 𝑒 in integrated absolute difference. For water facing chloride at 3.15 Å, the density change is 0.000107 𝑒 with the MACE-induced dipole, 0.00426 𝑒 with the prescribed DFT-induced dipole, and 0.01026 𝑒 with DFT at that same dipole. Thus the error involves both the field that induces chloride and the water response to the returning field. For Na+–Ar at 6 Å, the predicted solute multipoles are unchanged, while environment induction gives an interaction of −9.19 meV, close to the classical induction limit","source_metadata":{"date_field":"posted","keywords":[]}},{"id":"journals:10.1021/acs.jcim.6c02968","kind":"journals","source":"Journal of Chemical Information and Modeling","title":"Shake It Off: Abl\nKinase Regulatory Domain Dynamics\nExplain Conformational Changes upon Activation","url":"https://doi.org/10.1021/acs.jcim.6c02968","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.1021%2Facs.jcim.6c02968","date":"2026-11-19T00:00:00+00:00","timestamp":1795046400,"categories":["Free Energy & MD"],"topic_ids":["freeenergy"],"keywords":[],"matched_keywords":[],"matched_tags":["molecular_dynamics","biology","generic_chem"],"doi":"10.1021/acs.jcim.6c02968","external_id":null,"pdf_url":null,"code_url":null,"code_host":null,"authors":["Philipe Oliveira Fernandes","José Luis Muñoz Reyes","Pedro Tenório T. F. Leite","Diego Magno Martins","Antti Poso","Ina Pöhner","Adolfo Henrique Moraes","Vinicius G. Maltarollo","Thales Kronenberger"],"journal":"Journal of Chemical Information and Modeling","publisher":"ACS","impact_factor":5.3,"abstract":"Abl kinase is a major molecular target for chronic myeloid leukemia therapy. Its activity is tightly controlled by its regulatory module, composed of the SH3 and SH2 domains and the SH2–kinase linker. Although the assembled autoinhibited and extended active conformations have been structurally characterized, the molecular pathway connecting these states remains poorly understood. Simulations initiated from the end point structures failed to sample the complete transition, even when Gaussian accelerated molecular dynamics (GaMD) was employed to enhance conformational sampling. To overcome this limitation, we combined multiscale modeling with extensive molecular dynamics simulations and Markov state modeling to reconstruct the Abl disassembly pathway. The resulting kinetic landscape revealed two coupled conformational equilibria: a faster exchange between the canonical assembled state and an intermediate ensemble that captures the initial rotation of the regulatory module, and a slower transition toward the extended state, characterized by migration of the regulatory module atop the kinase domain N-lobe. The assembled ensemble is favored (∼88% of the equilibrium population), with the canonical autoinhibited arrangement alone accounting for approximately 60%, while early disassembly disrupts the E98–K294 and SH3–linker interactions and releases Y226 from its protective interactions, while Y70 remains accessible through the intrinsic dynamics of the SH3 domain. By revealing how these phosphorylation-sensitive residues become accessible during the earliest stages of disassembly, our model provides a mechanistic framework for Abl activation and identifies regulatory interfaces as potential targets for allosteric stabilization of the autoinhibited state.","source_metadata":{"short_name":"JCIM","source":"crossref","abstract_source":null,"abstract_quality":null}},{"id":"preprints:10.26434/chemrxiv.15012473/v1","kind":"preprints","source":"ChemRxiv","title":"Towards a Metrological Validation of PFAS Property Predictions by In-Context LLMs Across Four Architectures","url":"https://doi.org/10.26434/chemrxiv.15012473/v1","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.26434%2Fchemrxiv.15012473%2Fv1","date":"2026-11-19T00:00:00Z","timestamp":1795046400,"categories":["Property Prediction","LLMs & Agents"],"topic_ids":["property","llm"],"keywords":["QSPR","LLMs","LLM","DFT","dissociation constant","partition coefficient"],"matched_keywords":["QSPR","LLMs","LLM","DFT","dissociation constant","partition coefficient","chemical","molecular","molecules"],"matched_tags":["qsar_qspr","LLMs","dft","inhibition_constant","solubility","generic_chem"],"doi":"10.26434/chemrxiv.15012473/v1","external_id":"10.26434/chemrxiv.15012473/v1","pdf_url":null,"code_url":null,"code_host":null,"authors":["Carlos Gonzalez"],"journal":null,"publisher":null,"impact_factor":null,"abstract":"In recent years, the use of large-language models (LLMs) as prediction tools has become increasingly popular. Despite this popularity, the metrological validation of LLMs in the study of chemical problems has been somehow neglected to date. To illustrate this need in chemistry, we introduce a minimal, metrologically grounded framework for validating in-context learning (ICL)-based prediction in LLMs on molecular physicochemical properties with reliable estimation of uncertainty. Our framework integrates a multicycle testing protocol (Mc = 50 replicate cycles per model-property cell) with a six-metric bundle: central tendency (mean MAE over replicate cycles), reproducibility (interquartile range), regression performance (RMSE, and R2), information-theoretical lower bound on the amount of captured signal (𝐼\" in nats), and ordinal 2 performance (accuracy against the median in the dataset), plus three cross-property and crossmodel comparison metrics: NRMSE, Welch’s z-test, and Scaling Law Fit exponents. We use this framework to validate four LLMs, two dense (8B and 70B parameters; 1B = 109) and two mixture-of-experts (MoE; 17B active parameters with 109B and 400B in total), for predicting eight per- and polyfluoroalkyl substance (PFAS) properties: boiling point (BP), melting point (MP), vapor pressure (VP), Henry’s law constant (HLC), octanol-water partition coefficient (LogKow), aqueous solubility (LogS), acid dissociation constant (pKa), and critical micelle concentration (LogCMC). In this benchmark, the frontier MoE model (Maverick 400B parameters) achieves a mean validation NRMSE of 14.4 %, and the mid-scale MoE model (Scout, 109B parameters) outperforms our larger dense model (Llama-3.3-70B) despite having 4.1× fewer active parameters. This suggests that MoE routed capacity can substitute for activeparameter scale in chemical applications of ICL. When compared to published QSPR and DFT baselines, LLM-ICL beats these approaches on three of eight directly comparable properties (pKa by 43 %, LogKow by 31 %, HLC by 10 %) and trails behind on five (BP, MP, LogS, VP by 12 % to 75 %, and LogCMC by a factor of approx. 3.3), despite using approximately 32 incontext examples (32 for seven properties, 20 for HLC) versus the training sets comprising 102 to 105 molecules used in these approaches. We also discuss a validation-set anomaly in the case of LogKow, where our mid-scale dense model Llama-3.3-70B outperforms the frontier MoE model on all four validation metrics, possibly due to better coverage of the pretraining data. Permutation (shuffled-label) and information-access blinding controls show that genuine structure-response signal is present, but that the native-prompt accuracy depends substantially on pretrained, property-specific knowledge and familiar target scales. Overall, our results suggest a two-fold conclusion on the competitiveness of LLM-ICL: it succeeds in predicting chemical","source_metadata":{"date_field":"posted","keywords":[]}},{"id":"preprints:2611.25076v1","kind":"preprints","source":"arXiv","title":"ShEPhERD-score: Fast, adaptable three-dimensional alignment of molecular features","url":"https://arxiv.org/abs/2611.25076v1","detail_url":"/chemradar/article?u=https%3A%2F%2Farxiv.org%2Fabs%2F2611.25076v1","date":"2026-11-18T19:17:58Z","timestamp":1795029478,"categories":["Cheminformatics","Property Prediction","Docking & Screening"],"topic_ids":["cheminformatics","property","docking"],"keywords":["virtual screening","molecular features","molecular representations"],"matched_keywords":["virtual screening","machine learning","molecular features","molecular representations","molecular","ligand","protein"],"matched_tags":["virtual_screening","machine_learning","molecular_descriptor","generic_chem","biology"],"doi":null,"external_id":"2611.25076v1","pdf_url":"https://arxiv.org/pdf/2611.25076v1","code_url":"https://github.com/coleygroup/shepherd-score","code_host":"GitHub","authors":["Nate Somers","Kento A. Abeywardane","Connor W. Coley","Masha Elkin"],"journal":null,"publisher":null,"impact_factor":null,"abstract":"Three-dimensional (3D) molecular similarity scoring and alignment are widely employed in ligand-based virtual screening and molecular design, and are increasingly used for training and evaluating machine learning models. Existing open-source methods utilize various molecular representations spanning shape, electrostatics, pharmacophores, and hybrid representations, but their implementations as independent software packages with distinct application programming interfaces (APIs), optimizers, and limited GPU support make them difficult to compare, combine, or extend. Here, we develop ShEPhERD-score into a unified, modular platform for differentiable 3D molecular similarity scoring and alignment by combining diverse molecular representations with high-performance Numba CPU and Triton GPU kernels to support both large-scale library screening and pairwise similarity workflows. In matched comparisons with open-source 3D similarity tools on the DUDE-Z virtual screening benchmark, ShEPhERD-score performed comparably or better in eight of nine cases. Across the scoring modes in ShEPhERD-score, throughput ranges from 17,800 to 1.03 million alignments/second for ten-million-conformer libraries using a single NVIDIA L40S GPU. Moreover, we found that the most effective scoring mode varied by protein target, indicating that target-specific calibration against validation data may be valuable before large-scale screening and underscoring the need for a unified platform that provides diverse molecular representations. Together, these capabilities establish ShEPhERD-score as a single, extensible toolkit for rapid virtual screening and machine learning workflows.","source_metadata":{"categories":["q-bio.BM"],"code_status":"found","code_url":"https://github.com/coleygroup/shepherd-score"}},{"id":"preprints:2611.25052v1","kind":"preprints","source":"arXiv","title":"MAPPA: Mapping of Atoms by Progressive Pattern Assembly with Molecular Symmetry","url":"https://arxiv.org/abs/2611.25052v1","detail_url":"/chemradar/article?u=https%3A%2F%2Farxiv.org%2Fabs%2F2611.25052v1","date":"2026-11-18T19:02:04Z","timestamp":1795028524,"categories":["Reaction Informatics"],"topic_ids":["reaction"],"keywords":["Atom to atom mapping"],"matched_keywords":["Atom-to-atom mapping","Molecular","chemical"],"matched_tags":["atom_mapping","generic_chem"],"doi":null,"external_id":"2611.25052v1","pdf_url":"https://arxiv.org/pdf/2611.25052v1","code_url":"https://github.com/yunhzou/MAPPA-AAM","code_host":"GitHub","authors":["Yunheng Zou","Olalla Nieto Faza","Shifa Hussain","Varinia Bernales","Alan Aspuru-Guzik"],"journal":null,"publisher":null,"impact_factor":null,"abstract":"Atom-to-atom mapping traces atom identities between molecular endpoints, supporting reaction anal- ysis and pathway construction. Yet the same endpoints can admit multiple correspondences, and selecting a single answer can conceal alternatives with different chemical implications. We introduce MAPPA, a framework that efficiently discovers and retains alternative atom mappings. It uses shared molecular fragments and molecular symmetry to represent these alternatives compactly, allowing their bond changes to be examined and supporting subsequent chemistry-aware selection. MAPPA achieves the highest reference coverage among the evaluated methods on Golden. Standard benchmark runs re- cover 99.03% of Golden and 99.63% of EnzymeMap reference mappings, with matching and decoding together averaging 3.19 and 13.16 CPU seconds per reaction, respectively. Broader exploration raises Golden coverage to 99.51%. These results make broad exploration of mapping alternatives practical, providing a foundation for studying reaction ambiguity and incorporating chemical knowledge into mapping selection.","source_metadata":{"categories":["physics.chem-ph"],"code_status":"found","code_url":"https://github.com/yunhzou/MAPPA-AAM"}},{"id":"preprints:2611.25041v1","kind":"preprints","source":"arXiv","title":"Can Language Models Build Molecules in Three Dimensions?","url":"https://arxiv.org/abs/2611.25041v1","detail_url":"/chemradar/article?u=https%3A%2F%2Farxiv.org%2Fabs%2F2611.25041v1","date":"2026-11-18T19:00:08Z","timestamp":1795028408,"categories":[],"topic_ids":[],"keywords":["molecular shape"],"matched_keywords":["molecular shape","Molecules","Chemical","molecular","compounds","ligand"],"matched_tags":["shape_similarity","generic_chem"],"doi":null,"external_id":"2611.25041v1","pdf_url":"https://arxiv.org/pdf/2611.25041v1","code_url":"https://github.com/the-matter-lab/MolDraft-data","code_host":"GitHub","authors":["Hanyong Xu","Yunheng Zou","Olalla Nieto Faza","Juan Bernardo Pérez Sánchez","Aiwei Yin","Alán Aspuru-Guzik","Varinia Bernales"],"journal":null,"publisher":null,"impact_factor":null,"abstract":"Chemical drawings convey which atoms are bonded and how stereocentres are configured, but not molecular shape. Every simulation starts from three-dimensional coordinates, and building them from a drawing remains a manual step. We introduce MolDraft, a curated benchmark of 300 drawing-structure pairs spanning natural products, hypervalent compounds, and coordination complexes. Given an image and a short instruction, an AI system must generate a complete set of atomic coordinates, which are checked for molecular identity, stereochemical configuration, and coarse ligand geometry. We evaluate nine multimodal language models using the same agent setup. The share of molecules built correctly, the resolve rate, ranges from 25.0% to 76.0% and falls as molecules grow, most sharply above about 80 atoms. Only 29 of the 300 molecules are resolved by all nine models, and literature structures, mostly metal complexes, are the hardest source for every model. Language models can thus build many molecules in three dimensions, but not yet reliably reproduce the coordination chemistry of the research literature.","source_metadata":{"categories":["physics.chem-ph"],"code_status":"found","code_url":"https://github.com/the-matter-lab/MolDraft-data"}},{"id":"preprints:2611.24764v1","kind":"preprints","source":"arXiv","title":"Horizontal Flow Matching for Single-Cell Spatiotemporal Dynamics","url":"https://arxiv.org/abs/2611.24764v1","detail_url":"/chemradar/article?u=https%3A%2F%2Farxiv.org%2Fabs%2F2611.24764v1","date":"2026-11-18T17:16:53Z","timestamp":1795022213,"categories":[],"topic_ids":[],"keywords":["transformer","equivariant"],"matched_keywords":["transformer","equivariant","molecular"],"matched_tags":["transformer","equivariance","generic_chem"],"doi":null,"external_id":"2611.24764v1","pdf_url":"https://arxiv.org/pdf/2611.24764v1","code_url":null,"code_host":null,"authors":["Zihan Wang","Jianzhe Li","Zhenyi Zhang","Peijie Zhou"],"journal":null,"publisher":null,"impact_factor":null,"abstract":"Learning the spatiotemporal dynamics of cells from spatial transcriptomics requires connecting snapshots of different cells acquired in different coordinate frames. A rotation or translation of a section can appear as cell movement and distort the inferred correspondence. We introduce horizontal flow matching, which learns changes in tissue organization, molecular state, and cell abundance up to shared rigid motion. The dynamics follow a continuity--reaction equation whose action assigns no cost to shared rigid motion, and the resulting horizontal projector removes this component from spatial prediction errors during learning and from predicted velocities during generation. An equivalent endpoint problem jointly estimates correspondence and relative pose in Wasserstein--Fisher--Rao geometry and yields analytic paths whose velocity and growth serve as regression targets. Regressing these targets recovers the horizontal dynamics under ideal population conditions. Our model, HORIZON, uses an equivariant transformer to predict velocity and growth from the current tissue, so that a tissue can be generated continuously from a single initial snapshot. On a simulated tissue observed in perturbed coordinate frames, alignment improves reconstruction and projection reduces rigid drift. On four spatial transcriptomics time courses, HORIZON reconstructs tissue geometry, gene expression, and cell abundance more accurately than existing spatial and unbalanced trajectory methods, and interpolates held-out developmental stages.","source_metadata":{"categories":["cs.LG","cs.AI","q-bio.QM"],"code_status":"absent"}},{"id":"preprints:2611.24421v1","kind":"preprints","source":"arXiv","title":"Orbital-weighted learning for transition metal complexes","url":"https://arxiv.org/abs/2611.24421v1","detail_url":"/chemradar/article?u=https%3A%2F%2Farxiv.org%2Fabs%2F2611.24421v1","date":"2026-11-18T13:43:12Z","timestamp":1795009392,"categories":[],"topic_ids":[],"keywords":[],"matched_keywords":["neural network","machine learning","molecular","ligand"],"matched_tags":["neural_network","machine_learning","generic_chem"],"doi":null,"external_id":"2611.24421v1","pdf_url":"https://arxiv.org/pdf/2611.24421v1","code_url":null,"code_host":null,"authors":["Gabriel Klassen","Guido Falk von Rudorff"],"journal":null,"publisher":null,"impact_factor":null,"abstract":"Frontier orbital energies are key to the description of molecular light-matter interaction and pose a challenge both for electronic structure calculations and for machine learning approaches, especially for transition metal complexes (TMC). These are highly relevant since the gap between the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) is in the range of visible light. For TMCs, gap position and magnitude are determined by the metal center, its coordination and ligand identity. Since the systems exhibit a wide range of charge and spin states, frontier orbital energies can only be predicted with machine learning models that explicitly incorporate spin and charge. In this work, we show that a semi-empirical orbital population as a weight rather than a feature can increase efficiency for both physics-based and neural-network-based atomic representations. Because orbital populations are highly sensitive to both charge and spin, they offer a general pathway to render existing geometry-based machine learning representations useful for orbital energies. On TMC datasets, our approach reduces the required training data by up to 95 %, and improves physics-based representations to almost match the accuracy of pre-trained neural network representations at 12% of the inference compute cost. This greatly reduces the data and compute requirements to accurately model frontier orbitals in TMCs","source_metadata":{"categories":["physics.chem-ph"],"code_status":"absent"}},{"id":"preprints:2611.24148v1","kind":"preprints","source":"arXiv","title":"Machine learning interatomic potentials and orientational defects in polymers","url":"https://arxiv.org/abs/2611.24148v1","detail_url":"/chemradar/article?u=https%3A%2F%2Farxiv.org%2Fabs%2F2611.24148v1","date":"2026-11-18T09:54:37Z","timestamp":1794995677,"categories":["ML Potentials"],"topic_ids":["potentials"],"keywords":["MACE","density functional theory","DFT"],"matched_keywords":["Machine learning interatomic potentials","MACE","machine learning interatomic potential","Machine learning","density functional theory","DFT","molecular"],"matched_tags":["mlip_tools","machine_learning","dft","generic_chem"],"doi":"10.1007/978-94-024-2316-7_6","external_id":"2611.24148v1","pdf_url":"https://arxiv.org/pdf/2611.24148v1","code_url":null,"code_host":null,"authors":["Arnold Kiv","Volodymyr Soloviev","Andrii Tuzhykov","Andrii Bielinskyi","Taras Kavetskyy","Nataliia Hoivanovych","Ondrej Šauša","Jaroslav Rusnák","Helena Švajdlenková","Jolita Ostrauskaite","Mykola Slusarenko"],"journal":null,"publisher":null,"impact_factor":null,"abstract":"Photopolymerization of acrylated epoxidized soybean oil (AESO) is a convenient model of a self-organizing process in a soft, bio-based material: ultraviolet light starts a free-radical chain reaction, the growing network releases heat, and the mobility of molecular groups changes on several time scales at once. We follow this process from two sides. Experimentally, the real and imaginary parts of the dielectric permittivity of an AESO:VDM (1:0.5) mixture with 3 mol% of the photoinitiator DMPA were recorded at 500-5000 Hz under an intermittent on/off illumination protocol, and the resulting time series were analysed with a sliding-window Shannon entropy and the Grassberger-Procaccia correlation integral. Both measures show that short light pulses at the start of curing trigger strongly correlated, collective relaxation, which fades as the network forms, leaving only localized reorientations active. We describe these reorientations as orientational defects and quantify their energetics with the MACE-OFF23 machine learning interatomic potential, using relaxed dihedral scans of five rotational modes ($α$, $α/β$, $β$, $γ$, $δ$) of an AESO fragment. Compared with density functional theory (DFT) reference profiles, MACE reproduces the shape of all five profiles (Pearson $r=0.95$-$0.98$) and the hierarchy of barriers - from 2.9 eV for the constrained main-chain rotation, through 0.54-0.81 eV for the acrylate rotations, to 0.16-0.18 eV for the methyl and 0.06 eV for the hydroxyl rotation - while underestimating the barrier heights by 3-20%. A simple Arrhenius estimate shows that the acrylate rotations are the modes closest to the frequency window of the dielectric measurement.","source_metadata":{"categories":["cond-mat.mtrl-sci","cond-mat.soft","physics.chem-ph"],"code_status":"absent"}},{"id":"preprints:2611.23873v1","kind":"preprints","source":"arXiv","title":"Technical Note on Matlantis PFP v9: Extending the Element and Domain Coverage of the r2SCAN Training Data","url":"https://arxiv.org/abs/2611.23873v1","detail_url":"/chemradar/article?u=https%3A%2F%2Farxiv.org%2Fabs%2F2611.23873v1","date":"2026-11-18T05:25:35Z","timestamp":1794979535,"categories":["ML Potentials"],"topic_ids":["potentials"],"keywords":["density functional theory","DFT","PBE"],"matched_keywords":["machine learning interatomic potentials","machine learning","density functional theory","DFT","PBE","molecules"],"matched_tags":["mlip_tools","machine_learning","dft","generic_chem"],"doi":null,"external_id":"2611.23873v1","pdf_url":"https://arxiv.org/pdf/2611.23873v1","code_url":null,"code_host":null,"authors":["Daiki Shintani","Shigeru Iwase","Jun Oshiki","Kohei Shinohara","Nontawat Charoenphakdee","Makoto Hiramatsu","Yuta Tsuboi","Yusuke Asano","Hirotaka Akita","Yong-Bin Zhuang","So Takamoto","Chikashi Shinagawa"],"journal":null,"publisher":null,"impact_factor":null,"abstract":"Universal machine learning interatomic potentials (uMLIPs) make atomistic simulations possible at scales beyond the reach of density functional theory (DFT). Matlantis PFP is a uMLIP that treats a variety of domains, including crystals, molecules, and surfaces, with a single model. PFP v8 incorporated a training dataset computed with the r2SCAN functional, chosen for its good agreement with experiment, but the elements and domains covered by that dataset were limited. Here we present Matlantis PFP v9, in which the training data computed with the r2SCAN functional are extended to the lanthanides from Pr to Lu and to the elements from Po to Cm, for a total of 96 elements, and to additional domains such as clusters and adsorption structures. The formation energies of crystals containing the newly supported elements reproduce experimental standard enthalpies of formation with the same level of accuracy as crystals composed only of previously supported elements, and the adsorption energies of molecules on transition-metal surfaces have an overall error against experiment comparable to that of DFT-PBE, although chemisorbed species are overbound. On the same type of GPU (A100), the inference speed and the maximum system size exceed those of SevenNet-Omni, a public uMLIP trained on r2SCAN data: forces are evaluated three to four times faster, and five to ten times as many atoms can be handled. As a result, PFP-R2SCAN is now applicable to systems containing rare-earth and actinide elements and to adsorption on surfaces.","source_metadata":{"categories":["cond-mat.mtrl-sci","physics.comp-ph"],"code_status":"absent"}},{"id":"preprints:2611.23777v1","kind":"preprints","source":"arXiv","title":"STORM: A Lightweight Structure-Free Model for Binding Affinity Prediction","url":"https://arxiv.org/abs/2611.23777v1","detail_url":"/chemradar/article?u=https%3A%2F%2Farxiv.org%2Fabs%2F2611.23777v1","date":"2026-11-18T03:34:08Z","timestamp":1794972848,"categories":["Targets & Structures","Free Energy & MD"],"topic_ids":["targets","freeenergy"],"keywords":["Boltz 2","free energy perturbation","Binding Affinity"],"matched_keywords":["Boltz-2","free-energy perturbation","machine learning","protein","molecule","drug","Molecules","Binding Affinity"],"matched_tags":["ml_protein_ligand","free_energy_calcs","machine_learning","biology","generic_chem","bioactivity"],"doi":null,"external_id":"2611.23777v1","pdf_url":"https://arxiv.org/pdf/2611.23777v1","code_url":null,"code_host":null,"authors":["Ayush Pandit","Rohan V. Koodli","Wei Deng","Roxana Daneshjou","Ron O. Dror"],"journal":null,"publisher":null,"impact_factor":null,"abstract":"Accurately predicting a small molecule's binding affinity to a protein, a critical task in drug discovery, remains challenging. Physics-based approaches like free-energy perturbation (FEP) have very high computational costs and require expert setup. Recent machine learning--based cofolding models have reported accuracy approaching that of FEP---and at least one of them, Boltz-2, has already been widely adopted for affinity prediction---but they remain expensive for large computational screens and even more expensive for retraining on newly collected experimental data. To address these challenges, we introduce STORM: Similarity Transfer from Observed Reference Molecules. STORM is a fast, simple machine learning model that ensembles many small model fusions of diverse protein and molecule representations for binding affinity prediction. Trained on the same data, STORM matches or exceeds the accuracy of Boltz-2 on multiple evaluation datasets, while delivering a speed improvement of over five orders of magnitude for both inference and training. Our results also suggest opportunities to substantially improve affinity prediction accuracy by integrating historical experimental affinity data with structural and biophysical information.","source_metadata":{"categories":["q-bio.QM","cs.LG"],"code_status":"absent"}},{"id":"journals:10.1039/d6dd00558f","kind":"journals","source":"Digital Discovery","title":"Pocket-Based 3D Molecular Generation: Binding-Environment Representations and Pocket Adaptability","url":"https://doi.org/10.1039/d6dd00558f","detail_url":"/chemradar/article?u=https%3A%2F%2Fdoi.org%2F10.1039%2Fd6dd00558f","date":"2026-11-18T02:43:43Z","timestamp":1794969823,"categories":["Design de novo"],"topic_ids":["denovo"],"keywords":["Molecular Generation"],"matched_keywords":["Molecular Generation","Molecular","drug","ligands","protein"],"matched_tags":["de_novo_design","generic_chem","biology"],"doi":"10.1039/d6dd00558f","external_id":null,"pdf_url":null,"code_url":null,"code_host":null,"authors":["Bo Huang","Dongsheng Cao","Aiping Lu","Rong Qian","Tingfei Zhu","Jia-Cai Yi","Woruo Chen","Lei Wang","Yanjing Duan","Ling Zhang","Jin Liu"],"journal":"Digital Discovery","publisher":"RSC","impact_factor":5.6,"abstract":"Pocket-based 3D molecular generation has emerged as an important direction in structure-based drug design by enabling ligands to be generated directly within protein binding environments. Most existing methods rely on...","source_metadata":{"short_name":"Digital Discovery","source":"crossref","abstract_source":null,"abstract_quality":null}}],"total_article_count":3258,"filters":{"topic":"all","source":"all","days":2,"limit":null}}