The Memory Palace

Reading List

Like a todo-list but for knowledge. Ideally this page is auto-updated by the curating agent every morning as it reads the arXiv feeds…

02 Apr 2025

  • From GPCRs to Kinases: Automating the Path to Universal Pandemic Vaccine Development
    Asif Sumbul, Mukherjee Debaprasad
    Biological and Medicinal Chemistry on ChemRxiv
    2025-04-01
     The paper discusses the critical roles of GPCRs and kinases in viral infections and the urgent need for universal pandemic vaccines against strains like Covid-19, SARS, and MERS. It highlights the potential of AI and computational biology in automating vaccine development, particularly for broad-spectrum antiviral treatments.

  • µMap-FFPE: A High-Resolution Protein Proximity Labeling Platform for Formalin-Fixed Paraffin-Embedded Tissue Samples
    Noah B. Bissonnette, Marie E. Zamanis, Steven D. Knutson, Zane Boyer, Angelo Harris, Daniel Martin, Jacob B. Geri, Suzana Couto, Tahamtan Ahmadi, Anantharaman Muthuswamy, Mark Fereshteh, David W. C. MacMillan
    Biological and Medicinal Chemistry on ChemRxiv
    2025-04-01
     The paper presents µMap-FFPE, a novel protein proximity labeling platform for studying protein interactions in formalin-fixed paraffin-embedded (FFPE) tissue samples. The method aims to explore microscale interactions critical for understanding disease states, specifically comparing CD20’s interactome between healthy and cancerous tissues, overcoming challenges posed by formalin-fixation.

  • Four new bright members of the ZinPyr zinc fluorescence sensor family for live cell imaging
    Sabine Becker, Marisa Franziska Jakobs, Max Carlsson, Simon Wittmann, Jörg Fahrer
    Biological and Medicinal Chemistry on ChemRxiv
    2025-04-01
     The paper introduces four new members of the ZinPyr zinc fluorescence sensor family—ZP1(5-en), ZP1(6-en), ZP1(5-Me2en), and ZP1(6-Me2en)—developed for live cell imaging. These sensors exhibit low quantum yields in the zinc-free state but high turn-on and dynamic ranges. They successfully detect intracellular zinc and localize in lysosomes and nuclei.

  • Computational Insights into the Binding Mechanism of L6I with Keap1 Kelch Domain (5FNU): A Molecular Docking Approach
    Swaroop Kumar . P
    Biological and Medicinal Chemistry on ChemRxiv
    2025-04-01
     The paper investigates the binding mechanism of the L6I inhibitor with the Keap1 Kelch domain (PDB ID: 5FNU) using molecular docking. The study utilized AutoDock to analyze binding affinity, achieving -10.6 kcal/mol with a binding distance of 2.677 Å. Results indicate L6I’s potential to disrupt Keap1-Nrf2 interaction, suggesting further experimental validation is needed.

  • High-performance quantitative exposomics covering up to >230 toxicants and key biomarkers
    Yunyun Gu, Max Lennart Feuerstein, Caroline Helen Johnson, Benedikt Warth
    Biological and Medicinal Chemistry on ChemRxiv
    2025-04-01
     The paper presents a scalable LC-MS/MS workflow for quantifying over 230 toxicants and biomarkers in urine, plasma, and serum. The method demonstrated high sensitivity, robustness, and suitable extraction recovery, with over 130 biomarkers detected in 200 urine samples from pregnant women. This approach supports large-scale exposome-wide association studies (ExWAS) effectively.

  • Polyglycerol-based Lipids: A Next-Generation Alternative to PEG in Lipid Nanoparticles for Advanced Drug Delivery Systems
    Yara Ensminger, Rashmi Rashmi, Gideon Nölte, Michael Karimov, Ann-Cathrin Schmitt, David Diaz-Oviedo, Johannes Koebberling, Rainer Haag
    Biological and Medicinal Chemistry on ChemRxiv
    2025-04-01
     The paper presents a study on polyglycerol-based lipid nanoparticles (LNPs) as an alternative to PEG for drug delivery, addressing issues with anti-PEG antibodies. The researchers formulated LNPs with linear polyglycerol (lPG) and demonstrated that these LNPs had low binding to anti-PEG antibodies and comparable mRNA delivery efficacy to PEGylated LNPs in HepG2 cells.

  • In-situ Dynamic Imaging of Magnetosome Biomineralization in A Living Bacterium
    Jiandong Feng, Qin Juan, Yibo Yang, Zengrong Zhou, Pinlong Zhao, Shiyang Lyu, Wenjing Duan, Hongzhen Bai, Jinhua Li
    Biological and Medicinal Chemistry on ChemRxiv
    2025-04-01
     The paper presents in-situ dynamic magnetic imaging of magnetosome biomineralization in living magnetotactic bacteria using Quantum-enabled Chemical Operando Microscopy. The study reveals a novel ‘retarded logistic type dynamics’ for magnetism development during magnetosome chain formation and elucidates the growth and assembly mechanisms at a sub-bacterium level, advancing the understanding of biomineralization in living organisms.

  • Structure-based identification of the first non-covalent thioredoxin reductase inhibitor with proven ADMET suitability
    Luisa Ronga, Giuseppe Felice Mangiatordi, Lamanna Giuseppe, Augello Giuseppa, Tesauro Diego, Silvestri Ilaria, Azzolina Antonina, Melchiorre Cervello, Saviano Michele
    Biological and Medicinal Chemistry on ChemRxiv
    2025-04-01
     The study identifies C55, the first non-covalent thioredoxin reductase 1 (TrxR1) inhibitor with established ADMET safety. Using a computational approach, over 90,000 compounds were screened, revealing C55’s micromolar-range inhibitory effect on TrxR1 across various cancer cell lines (HepG2, Huh7, MCF-7, MDA-MB-231), paving the way for anticancer drug development.

01 Apr 2025

  • Achieving excellent optical molecules by screening superalkali-doped cyclo[2n]carbon, M3O@C2n (M = Li, Na, and K, n = 5-10)
    Zeyu Liu, Wenwen Zhao, Jiaojiao Wang, Xiufen Yan, Tian Lu
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-31
     The paper explores nonlinear optical (NLO) molecules formed by superalkali-doped cyclo[2n]carbon (M3O@C2n, M = Li, Na, K; n = 5-10). Using time-dependent density functional theory, it finds that the polarizability and hyperpolarizability increase with alkali metal size. Specifically, Li3O@C20 shows exceptional properties, including a red-shifted absorption spectrum and potential for deep-ultraviolet NLO applications.

  • QRCI: Quantitative Ring Complexity Index for Evaluating Molecular Structure and Chemical Diversity
    Jianmin Wang, Kai Xu, Tengfei Ma, Xudong Zhang, Pengsen Ma, Chunyan Li, Weiran Huang, Meng Wang, Xiaojun Yao, Tao Song, Xiangxiang Zeng
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-31
     The paper introduces the Quantitative Ring Complexity Index (QRCI), which enhances traditional ring complexity assessments by incorporating ring diversity, topological complexity, and macrocyclic features. QRCI can be computed without 3D data and correlates with synthetic accessibility, aiding in evaluating molecular structures and chemical diversity for drug discovery amidst vast chemical spaces.

29 Mar 2025

  • Identification of Betulinic Acid Derivatives as Inverse Agonists of RAR-related Orphan Receptor Gamma (RORb3)
    Ya Chen, Patrik F. Schwarz, Alexander F. Perhal, Famke Guder, Jorge Enrique Hernández González, Kerrin Janssen, Ece Sağiroğlu, Ammar Tahir, Johannes Kirchmair, Natacha Rochel, Verena M. Dirsch
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-28
     The paper identifies novel betulinic acid derivatives as ROR-b3 inverse agonists, crucial for inhibiting Th17 cell differentiation linked to autoimmune diseases. A key compound exhibited IC50 values of 0.4 M and 0.6 M in ROR-b3 assays. RT-qPCR confirmed gene down-regulation, with structure-activity relationships explored through molecular docking and mutagenesis.

  • Cryo-EM exposes diverse polymorphism in IAPP mutants to guide rational design of novel therapeutics
    Dylan Valli, Saik Ann Ooi, Mikołaj Ignacy Kuska, Helena Marí, Himanshu Chaudhary, Weixiao Yuan Wahlgren, Sebastian Westenhoff, Alesia A Tietze, Anna Novials, Joan-Marc Servitja, Michał Maj
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-28
     The paper investigates IAPP mutants for type 2 diabetes therapies using cryo-electron microscopy to reveal six polymorphs from single-point mutations A25P, S28P, and S29P. Notably, the A25P variant formed a unique trimeric structure. Structures guided the design of novel non-amyloidogenic IAPP mutants, showcasing the potential of structure-based therapeutic development against amyloid diseases.

  • Ligand binding into a cryptic pocket in Escherichia coli DsbA inhibits the enzymatic activity in vitro
    Martin Scanlon, Biswaranjan Mohanty, Wesam Alwan, Menachem Gunzburg, Olga Ilyichova, Martin Williams, Gaurav Sharma, Geqing Wang, Begona Heras, Bryn Fenwick, Peter Wright, Pramodh Vallurapalli, Bradley Doak, Ben Capuano
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-28
     The paper identifies a cryptic ligand binding pocket in Escherichia coli DsbA using nuclear magnetic resonance (NMR) relaxation-dispersion data. It describes a dynamics-driven fragment discovery method and reveals that compounds bind to the pocket with slow kinetics, inhibiting DsbA’s activity in vitro. This suggests the pocket’s potential for developing new anti-virulence agents targeting EcDsbA.

  • Integrative Design of ADAMTS Partial Agonists within a Multi-Hallmark Therapeutic Framework: The Beginning of Paving the Way Toward a Functional Cure for Neurodegenerative Disorders
    David Ferguson, MRSB MRSC
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-28
     The paper discusses a multi-hallmark therapeutic approach to neurodegenerative disorders, focusing on ADAMTS metalloproteinases. It emphasizes bioisosteric design in developing partial agonists to control ECM degradation and reduce neuroinflammation. The integration of immune modulation and drug delivery systems is highlighted as enhancing treatment efficacy, aiming to restore functional neural networks in diseases like Alzheimer’s and Parkinson’s.

  • Smart sensor particles for intracellular tracing of reactive oxygen species by transmission electron microscopy
    Stephanie Hoeppener, Maren T. Kuchenbrod-Röther, Paul M. Jordan, Phillip Dahlke, Alicia De San Luis Gonzalez, Steffi Stumpf, Oliver Werz
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-28
     The paper presents a novel sensor particle system for detecting reactive oxygen species (ROS) within cells using Transmission Electron Microscopy (TEM). The method involves linking cerium chloride to poly(2-vinylpyridine) particles, reducing cytotoxicity and expanding cultivation media options. This approach enhances existing STAINING protocols for ROS detection.

28 Mar 2025

  • Binding modes of the KRAS(G12C) inhibitors GDC-6036 and LY3537982 revealed by all atom molecular dynamics simulations
    Tatu, Pantsar; Renne, Leini; Kari, Kopra; Jonas, Kapp
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-27
     The paper reveals the binding modes of KRAS(G12C) inhibitors GDC-6036 and LY3537982 using all-atom molecular dynamics simulations, totaling 200 microseconds. Experimental assays show high affinity for both inhibitors towards KRAS(G12C), but efficacy is reduced by certain co-mutations. The findings support the use of microsecond simulations for predicting binding interactions, aligning with experimental data.

  • Investigating errors in alchemical free energy predictions using random forest models and GaMD
    Skanda, Sastry; Michael, Kim
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-27
     The paper investigates alchemical free energy (ΔG) prediction errors using random forest models and GaMD for antibody-antigen complexes. It highlights that current predictions, while sufficient for screening, fail to assess critical post-translational modifications that can significantly impair binding. Key metrics include a ΔG threshold of +0.5 kcal/mol, indicating a 50% loss in dissociation constant (Kd).

27 Mar 2025

  • Leveraging High-Spin DFT Features for Prediction of Spin State Gaps in 3d Transition Metal Complexes
    Anuj Kumar, Ray; Sandeep , Nagar; Girish, Varma; U Deva, Priyakumar; Ankan, Paul
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-26
     The paper presents a machine learning approach to predict spin state energetics in 3d transition metal complexes using features from a single high-spin DFT calculation. Trained on 1434 SSE values from 934 complexes, the method retains accuracy while bypassing multi-reference optimizations, demonstrating transferability to more complex ligands. Key features include atomic energy levels and d-orbital eigenvalues.

  • Molecular aromaticity: a quantum phenomenon
    Miquel, Solà; Dariusz W. , Szczepanik
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-26
     The paper re-evaluates the classification of cyclo[18]carbon, a molecule identified in 2019, questioning its aromaticity based on quantum mechanics principles. The analysis suggests it should be considered non-aromatic. No specific datasets or experimental methods are mentioned in the abstract.

  • Predicting Catalytic Activity for CH4 Combustion on Pd-exchanged Zeolite Catalysts Using Automated Reaction Route Mapping
    Shunsaku, Yasumura; Shiho, Sakuma; Kenichiro, Saita; Tetsuya, Taketsugu; Masaru, Ogura
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-26
     The paper discusses predicting CH4 combustion catalytic activity using Pd-exchanged zeolites through automated reaction route mapping and neural network potential. Key findings include identifying intermediate species and evaluating activation energies, with Pd-MOR showing the highest catalytic activity. The research highlights the effectiveness of NNP in rationalizing catalyst design and pre-screening.

  • Generative deep learning for de novo drug design – a chemical space odyssey
    Rıza, Özçelik; Helena, Brinkmann; Emanuele, Criscuolo; Francesca, Grisoni
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-26
     The paper explores the use of generative deep learning in drug design, addressing challenges in molecule generation and evaluation. It highlights the importance of balancing chemical diversity, synthesizability, and bioactivity. The authors discuss current methods, optimization, and evaluation protocols, outlining future directions for utilizing generative models in exploring chemical space for novel drug development.

  • PolyMetriX: An Ecosystem for Digital Polymer Chemistry
    Kevin Maik, Jablonka; Sreekanth, Kunchapu
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-26
     The paper introduces PolyMetriX, an open-source Python library that standardizes and improves the digital polymer chemistry workflow. It provides curated polymer property datasets, hierarchical featurization techniques, and polymer-specific data splitting strategies. These enhancements lead to better model performance and reproducibility in polymer design, addressing issues of comparability in the field.

  • Active learning FEP using 3D-QSAR for prioritizing bioisosteres in medicinal chemistry
    Venkata Krishnan, Ramaswamy; Matthew, Habgood; Mark, Mackey
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-26
     The paper presents an active learning workflow combining 3D-QSAR and relative binding free energy calculations to prioritize bioisosteres in drug discovery. Demonstrated on a human aldose reductase case, this method efficiently identifies potent bioisosteric replacements from large datasets, achieving significant results with modest computational resources.

  • Machine learning-guided materials and system co-design for high-pressure hydrogen compression
    Matthew, Witman; Brendan, Davis; Vitalie, Stavila; Terry, Johnson
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-26
     The paper discusses enhancing hydrogen compression technology using machine learning for co-design of materials and systems. It presents a method for optimizing low-stability metal hydrides for high-pressure applications, validated through pressure-composition-temperature isotherms, and models compressor efficiency, predicting operational efficiencies for 2-stage and 3-stage compressors.

26 Mar 2025

  • Spatiotemporal Control of Formation of Dynamic Protein Fibre Networks via Photophysical Effects of a Focused Laser Beam
    Hiroshi Y. Yoshikawa, Ren Shirata, Takuya Takeshige, Fumika Kiryu, Kei Takano, Shuma Matsumoto, Natsumi Sawada, Chi-Shiun Wu, Yang-Hsin Shih, Hiromasa Niinomi, Takahisa Matsuzaki, Seiichiro Nakabayashi, Teruki Sugiyama, Ryuzo Kawamura
    https://chemrxiv.org/engage/rss/chemrxiv?categoryId=605c72ef153207001f6470d0
    2025-03-25
     The paper demonstrates a method for controlling the formation of dynamic microtubule networks using focused laser irradiation, which accumulates tubulin proteins at laser foci. The resulting networks exhibit dynamic behaviors like motion and cilia-like beating, driven by motor proteins. This technique relies on laser trapping and heat generation, enabling complex network fabrication without photochemical reactions, with potential applications in bioengineering.
  • Discovery of a Dual-Target Natural Compound for Rheumatoid Arthritis: High-Throughput Screening and Molecular Dynamics of a COX-2/JAK1 Inhibitor
    Dhrubajyoti Maji
    https://chemrxiv.org/engage/rss/chemrxiv?categoryId=605c72ef153207001f6470d0
    2025-03-25
     The paper reports the discovery of a dual-target compound, F3139-1037, for rheumatoid arthritis through high-throughput screening of 3666 natural compounds. This compound demonstrates superior binding affinity and inhibition for COX-2 and JAK1 compared to standard drugs. Molecular dynamics simulations confirm its stability, suggesting F3139-1037 as a promising candidate for multi-target RA therapies.
  • Counterintuitive Photochemistry of an Isolated Acridinyl Radical: ConPET via Preassembly Despite a Long-Lived Excited State
    Daniel Scott, Samuel Horsewill, Katherine Sharrock, P9r Feh9r, Jack Woolley, Imre P9pai
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-25
     The paper investigates the photochemistry of an isolated acridinyl radical in consecutive photoinduced electron transfer (conPET). Using photoreactivity, photoluminescence, and transient absorption techniques, it reveals that preassembly of the catalyst and substrate occurs prior to excitation, challenging previous assumptions about the role of a long-lived fluorescent excited state in reactivity.
  • Capturing Variability in Material Property Predictions for Plastics Recycling via Machine Learning
    Marcin Pietrasik, Anna Wilbik, Yannick Damoiseaux, Tessa Derks, Emery Karambiri, Shirley de Koster, Daniel van de Velde, Kim Ragaert, Sin Yong Teng
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-25
     The paper presents a framework utilizing interval-based machine learning to predict material properties in plastics recycling, addressing the challenges of heterogeneous polyolefin blends and stochastic measurements. The study involves a regressor for point estimation and an interval predictor, with empirical evaluations on a real-world dataset showing improved model interpretation and uncertainty prediction in mechanical recycling processes.
  • Circularly polarized luminescence modulation in europium(III) complexes based on camphor-fused bis(pyrazolyl)pyridine ligands
    Narcis Avarvari, Justin Cado, Alexandre Abherv9, Maxime Grasser, Nicolas Vanthuyne, Francesco Zinna, Lorenzo Di Bari, Boris Le Guennic
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-25
     The paper discusses europium(III) complexes formed using chiral camphor-fused bis(pyrazolyl)pyridine ligands. The ligands were fully characterized using single crystal X-ray diffraction. Key findings include strong circularly polarized luminescence (CPL) with glum factors up to ±0.2 and an ability to invert emission polarization between regioisomeric ligands, facilitating CPL modulation.
  • Graph2Mat: Universal graph to matrix conversion for electron density prediction
    Arghya Bhowmik, Pol Febrer, Peter J9rgensen, Miguel Pruneda, Alberto Garcia, Pablo Ordejon
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-25
     The paper introduces Graph2Mat, a method for converting molecular graphs into equivariant matrices to predict electron densities efficiently. It demonstrates state-of-the-art performance using significantly smaller datasets than traditional grid-based methods. Testing on QM9 molecules in DFT calculations yielded a 40% reduction in self-consistent field iterations, enabling new uncertainty measures for practical applications in machine learning-accelerated DFT methodologies.
  • Challenges and Opportunities for Machine Learning Potentials in Transition Path Sampling: Alanine Dipeptide and Azobenzene Studies
    Nikita Fedik, Wei Li, Nicholas Lubbers, Benjamin Nebgen, Sergei Tretiak, Ying Wai Li
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-25
     The paper discusses using machine learning (ML) potentials in transition path sampling (TPS) for the studies of alanine dipeptide and azobenzene. Utilizing the HIP-NN-TS and ANI-1x ML frameworks trained on a 5 million HCNO structure dataset, the study achieved near-quantum accuracy in conformational searches. Key findings include improved accuracy through active learning and challenges in azobenzene isomerization.
  • Data-Driven Many-Body Simulations of Biomolecules with CCSD(T) Accuracy: I Polyalanine in the Gas Phase
    Francesco Paesani, Ruihan Zhou, Ethan F. Bull-Vulpe, Yuanhui Pan
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-25
     The paper presents a data-driven many-body method, MB-nrg, achieving CCSD(T) accuracy for simulating polyalanine in the gas phase. By utilizing modular potential energy functions, the approach reproduces energies and free-energy landscapes more accurately than traditional force fields. It highlights improvements in capturing structural motifs and facilitates flexible sampling in longer peptides, advancing accurate simulations of proteins.
  • Stay Together or Split Up? Investigating Selective Adsorption of Carbon Dioxide and Acetylene in Anion-Pillared Microporous MOFs
    Mikko Karttunen, Yining Huang, Tahereh Azizivahed, Anupom Roy, Lucier Bryan, Shoushun Chen, Chen-Xi Liang
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-25
     The paper investigates selective adsorption of carbon dioxide and acetylene in anion-pillared microporous MOFs, specifically SIFSIX-1-Cu and SIFSIX-3-Cu. Methods include solid-state NMR spectroscopy and density functional theory, revealing SIFSIX-1-Cu favors acetylene while SIFSIX-3-Cu adsorbs both gases. Results enhance understanding of gas separation mechanisms and inform the design of selective adsorption materials.

25 Mar 2025

  • From GPCRs to Kinases: Automating the Path to Universal Pandemic Vaccine Development
    Sumbul , Asif; Debaprasad, Mukherjee
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-24
     The paper discusses automating the development of universal pandemic vaccines by targeting G protein-coupled receptors (GPCRs) and kinases involved in viral entry and replication. It emphasizes the need for such vaccines against varied strains, notably SARS-CoV-2. The authors highlight AI, machine learning, and computational biology as key methods for enhancing vaccine development.
  • Modulation of A3-40 and A4-40 co-assembly by Zinc: getting closer to the biological reality
    Enrico , Falcone; Lucie, de Cremoux; Christelle, Hureau; David, Schmitt; Wojciech, Bal; Ewelina , Stefaniak; Marta D., Wi śniewska; Nicolas, Vitale
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-24
     The paper investigates the effect of Zn(II) ions on the self- and co-assembly of amyloid-b2 peptides A1-40 and A4-40, using pH-dependent X-ray absorption and NMR spectroscopy. It finds that Zn(II) concentration influences assembly behavior, promoting potentially toxic aggregates. A4-16 has a lower Zn affinity than A1-16, highlighting the significance of Zn in Alzheimer’s disease pathology.
  • Complex structure-free compound-protein interaction prediction for mitigating activity cliff-induced discrepancies and integrated bioactivity learning
    Yaowen, Gu; Song, Xia; Qi, Ouyang; Yingkai, Zhang
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-24
     The paper presents CPI2M, a benchmark dataset with about 2 million bioactivity endpoints and activity cliff annotations, addressing challenges in compound-protein interaction (CPI) prediction. The authors introduce GGAP-CPI, a structure-free deep learning model that improves prediction accuracy by mitigating the effects of activity cliffs. GGAP-CPI outperforms several baselines across various scenarios in drug screening.
  • Two-way mechanism of designer biomolecular condensate catalysts
    Ayala, Lampel; Hao, Dong; Tlalit , Massarano; Yuqin , Yang; Avigail, Baruch Leshem; Ori, Eran; Xiaoyu, Wang
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-24
     The paper investigates designer biomolecular condensates formed by histidine peptides that catalyze ester hydrolysis through two mechanisms: Zn2+-dependent coordination and hydrogen bonds in the absence of Zn2+. Utilizing computational and experimental methods, the study highlights the catalytic pathways of these condensates, suggesting their potential in green chemistry and advanced materials.
  • How to generalize machine learning models to both canonical and non-canonical peptides
    Raúl, Fernández-Díaz; Rodrigo, Ochoa; Thanh Lam, Hoang; Vanessa, Lopez; Denis, Shields
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-24
     The paper examines methods for generalizing machine learning models to both canonical and non-canonical peptides. It highlights the superiority of chemical fingerprint-based similarity over traditional metrics in dataset partitioning. Deep-learned embeddings from Chemical Language Models outperform other models, but combining canonical datasets with non-canonical peptides enhances generalization. All supplementary materials are available on GitHub.
  • Machine learning transition state geometries and applications in reaction property prediction
    Isaac W., Beaglehole; Miles J., Pemberton; Elliot H. E., Farrar; Matthew N., Grayson
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-24
     This paper discusses the use of machine learning (ML) for predicting transition state (TS) geometries, aiming to reduce reliance on expensive quantum mechanical calculations. It reviews ML methods, their applications in predicting reaction properties, and the importance of accurate TS data. The paper also evaluates the limitations and challenges of current TS prediction methods.

22 Mar 2025

  • A 3D-Bioprinted Dermal-Like Scaffold Incorporating Fibroblasts and DRG Neurons to investigate Peripheral Nerve Regeneration
    Simone Bonetti, Francesco Formaggio, Emanuela Saracino, Eva Clemente, Marianna Barbalinardo, Franco Corticelli, Sara Buoso
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-21
     The paper presents a 3D-bioprinted fibroblast/DRG neuron co-culture scaffold to study peripheral nerve regeneration. The study optimized bioprinting parameters, achieving a biocompatible construct that supports cell growth. Results showed high viability of fibroblasts and enhanced neurite outgrowth in DRG neurons, emphasizing fibroblasts’ role in axonal regeneration, and offering a platform for neuropathic pain therapies.

  • AlphaFold2-RAVE: Protein Ensemble Generation with Physics-Based Sampling
    Da Teng, Vanessa J. Meraz, Akashnathan Aranganathan, Xinyu Gu, Pratyush Tiwary
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-21
     The paper presents AlphaFold2-RAVE (af2rave), an open-source Python tool that combines machine learning and physics-based sampling to generate alternative protein conformations. It utilizes reduced MSA from AlphaFold2 and molecular dynamics simulations, validating on E. coli adenosine kinase and human DDR1 kinase. af2rave offers efficient sampling, outperforming traditional methods, making it a valuable resource for structural biology and drug discovery.

  • Detecting apelin isoforms in human plasma using LC-MS/MS
    Maxwell Zeigler, Rozenn N. Lemaitre, Nona Sotoodehnia, Rheem Totah
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-21
     The paper presents a novel ultra-high-performance liquid chromatography-tandem mass spectrometry (LC-MS/MS) method for quantifying apelin isoforms (apelin-12, apelin-13, [Pyr1]-apelin-13, and apelin-17) in human plasma. This method achieves picomolar sensitivity and less than 20% variation, enabling the detection of apelin peptides as potential biomarkers for various health conditions.

  • SubTuner: a Physics-Guided Computational Tool for Modifying Enzymatic Substrate Preference and Its Application to Anion Methyltransferases
    Qianzhen Shao, Asher C. Hollenbeak, Yaoyukun Jiang, Brian O. Bachmann, Zhongyue J. Yang, Xinchun Ran
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-21
     The paper presents SubTuner, a physics-guided computational tool for modifying enzymatic substrate preference, particularly for anion methyltransferases. It utilizes three datasets with 190 and 600 mutants to demonstrate its speed, accuracy, and predictivity in identifying beneficial enzyme mutations for non-native substrates. SubTuner outperformed other engineering tools, highlighting its potential for enzyme engineering applications.

  • Hierarchical incremental learning deciphers multi-component materials
    Haoyuan Li, Hanyin Zhang, Nan Lin, Austin Evans, Tonghui Wang, Saied Pratik, Jean-Luc Bredas
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-21
     The paper presents HiDiscover, a hierarchical incremental learning protocol for analyzing multi-component materials. It enhances the understanding of Li-ion transport, gas adsorption, and molecular packing by revealing microscopic features not easily detectable with traditional simulations. This method improves reliability across different materials and processes, showcasing the effectiveness of incremental learning in materials design.

  • Active learning meets metadynamics: Automated workflow for reactive machine learning potentials
    Fernanda Duarte, Veronika Juraskova, Tristan Johnston-Wood, Hanwen Zhang, Valdas Vitartas
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-21
     The paper presents an automated workflow that integrates active learning and well-tempered metadynamics to enhance machine learning-based potentials (MLPs) for reaction modeling. It utilizes data-efficient architectures and demonstrates effectiveness through various organic reactions, achieving accurate MLPs without prior transition state knowledge, thus expanding applicability in reactive systems.

  • Mapping proton-coupled electron transfer with real space dimensions
    Adam Srut, Martin Diefenbach, Marvin L. Kronenberger, Benjamin J. Lear, Vera Krewald
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-21
     The paper presents a computational method for mapping proton-coupled electron transfer (PCET) onto real space dimensions, generating potential energy surfaces. It introduces a new criterion for ‘concerted’ proton-electron transfer by identifying proton and electron transfer coordinates, enhancing the understanding of their contributions and distinguishing between concerted and coupled transfer processes.

  • Quantifying Multidimensional Effects of Physicochemical Parameters on PFAS Adsorption Using a Hybrid Response Surface Methodology-Machine Learning Approach
    Renzun Zhao, Harsh V. Patel, Jazmin Green, Hyo-Shin (John) Park, Stephanie Luster-Teasley Pass
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-21
     The study develops hybrid response surface methodology-machine learning models to optimize PFAS adsorption. It utilizes a comprehensive dataset of experimental adsorption data, incorporating parameters like pH and molecular properties. The meta-learning HOP-RSM-GB model achieved near-perfect accuracy (R² = 1.00). Findings reveal optimal conditions for adsorption and emphasize the benefits of combining statistical and machine learning techniques in environmental remediation.

  • Toward Predictive Data-Driven Atomistic Modeling of Electrocatalyst Stability and Surface Reconstruction
    Jiayu Peng
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-21
     The paper evaluates methods for modeling electrocatalyst stability and surface reconstruction, addressing challenges in understanding dissolution and restructuring. It examines classical and data-driven approaches, including first-principles simulations and machine learning techniques, highlighting their strengths and weaknesses. The findings aim to enhance predictive modeling for optimizing electrocatalysts under harsh conditions.

  • CoRE MOF DB: a curated experimental metal-organic framework database with machine-learned properties for integrated material-process screening
    Yongchul Chung, Guobin Zhao, Logan Brabson, Saumil Chheda, Ju Huang, Thang Pham, Lionel Zoubritzky, Kunhuan Liu, Gianmarco Terrones, Kenji Mochida, Sunghyun Yoon, Haewon Kim, Prerna Prerna, Mohamad Moosavi, François-Xavier Coudert, Heather Kulik, David Sholl, Ilja Siepmann, Randall Snurr, Maciej Haranczyk
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-21
     The paper introduces an updated CoRE MOF database, featuring a curated collection of computation-ready metal-organic frameworks (MOFs) for efficient materials discovery. It includes machine-learned properties and employs Gibbs-Ensemble Monte Carlo simulations for hydrophobicity classification. The dataset was used for integrated material-process screening, identifying several MOF candidates that may exceed the performance of CALF-20 in carbon capture scenarios.

  • Machine learning reveals structural characteristics of stereochemistry-specific interdigitation of synthetic monomycoloyl glycerol analogs
    Suvi Heinonen, Artturi Koivuniemi, Matthew Davies, Mikko Karttunen, Camilla Foged, Alex Bunker
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-21
     The paper investigates the structural characteristics of synthetic monomycoloyl glycerol (MMG) analogs using machine learning (ML) and molecular dynamics (MD) simulations. It compares two analogs, MMG-1 and MMG-6, analyzing their phase behavior across three membrane states. Results indicate that interdigitation propensity correlates with hydrocarbon chain ordering, enhancing understanding of lipid-based formulations.

  • Quantifying design principles for light-emitting materials with inverted singlet-triplet energy gaps
    Laura McCaslin, Varun Rishi, Ali Abou Taka, Hrant Hratchian
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-21
     The paper analyzes design principles for organic light-emitting materials with inverted singlet-triplet energy gaps (ΔEST). Using a benchmark dataset of 15 heptazine-based molecules (HEPTA-INVEST15) and an expanded set (HEPTA-INVEST46), it finds strong linear correlations (R² > 0.94) between molecular structure and ΔEST. The study emphasizes the importance of strategic functionalization for efficient design of high-performance organic emitters.

  • A Minimalistic Deep Graph Learning Approach for Protein-Ligand Binding Affinity: One Step Towards Generalization
    Norberto Sánchez-Cruz, Ulises Rojas-Castañeda, Gabriel A. Argüelles-Arjona, Víctor H. Ramón-Cetina
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-21
     The paper presents ECIF-GCN, a minimalist deep graph learning model for predicting protein-ligand binding affinity, trained on the LP-PDBbind dataset. Utilizing Graph Convolutional Networks, it achieved a test RMSE of 1.52, outperforming more complex models. This demonstrates that effective prediction can be attained without overparameterization, emphasizing the benefits of simpler architectures in molecular modeling.

21 Mar 2025

  • Domain adaptable language modeling of chemical compounds identifies potent pathoblockers for Pseudomonas aeruginosa
    Georgios, Kallergis; Ehsannedin, Asgari; Martin, Empting; Anna, Hirsch; Alice, McHardy; Frank, Klawonn
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-20
     The paper presents ChemLM, a transformer-based language model for predicting molecular properties and identifying potent pathoblockers for Pseudomonas aeruginosa. It utilizes self-supervised domain adaptation, achieving superior performance on benchmark datasets. ChemLM showed significantly higher accuracy in identifying pathoblockers compared to existing methods, confirming its effectiveness and applicability in chemical property prediction.

  • Tuning Antiaromaticity through meso-Substituent Orientation in Core-Modified Isophlorins
    Ken-ichi, Yamashita; Maika, Isoda; Haruna, Sugimura; Yusuke, Honda
    Organic Chemistry on ChemRxiv
    2025-03-20
     This study explores the impact of meso-substituents on the antiaromaticity of stable dithiadioxaisophlorins (S2O2Iphs). It synthesized two new derivatives and utilized techniques like ¹H NMR and UV/vis spectroscopy alongside computational methods. Results reveal that steric bulk influences the tilt angle of substituents, affecting antiaromaticity, which aids in designing materials with adjustable electronic properties.

  • Bulky Bipyridine-Type Ligand-Enabled para-Selective C–H Borylation of Aromatic Compounds
    Yoichiro, Kuninobu; Taisei, Enta; Genki, Yoshino
    Organic Chemistry on ChemRxiv
    2025-03-20
     The paper presents a method for para-selective C–H borylation of aromatic compounds using iridium catalysts with bulky bipyridine-type ligands. By incorporating bulky substituents at specific positions, the method enhances para-selectivity, even with small substituents. Results indicate improved selectivity over traditional methods.

  • Electrochemical Oxidation of Benzyl Alcohols via Hydrogen Atom Transfer Mediated by 2,2,2-Trifluoroethanol
    Takahiro, Kawajiri; Masahiro, Hosoya; Satoshi, Goda; Eisuke, Sato; Seiji, Suga
    Organic Chemistry on ChemRxiv
    2025-03-20
     The paper presents a novel electrochemical oxidation method for benzyl alcohols using 2,2,2-trifluoroethanol as a hydrogen atom transfer mediator. Density-functional theory, cyclic voltammetry, and constant potential electrolysis validate the HAT mechanism, enabling oxidation of challenging substrates. The resulting carbonyl compounds can be further functionalized in an electrochemical one-pot process, enhancing their synthetic utility.

  • Direct Site-selective Deoxygenation of Benzylalcohol Derivatives
    Yoshihiro, Sato; Satoshi, Yagi; Kazuhiro, Morisaki
    Organic Chemistry on ChemRxiv
    2025-03-20
     The paper describes a straightforward method for the direct and site-selective deoxygenation of benzyl alcohol derivatives using Hantzsch ester, Cs₂CO₃, and a photocatalyst under photoirradiation. Mechanistic studies indicate that in situ reversible acylative activation of hydroxy groups is crucial for selectivity, providing a practical strategy for transforming polyols.

  • Structure-Based Modeling of Environment-dependent Protonation States Across LNP Formulations with Atomistic CpHMD
    Severin T., Schneebeli; Kyle J., Colston; Santiago C., Monsalve
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-20
     The paper introduces a scalable continuous constant pH molecular dynamics (CpHMD) model to study environment-dependent protonation states in lipid nanoparticle (LNP) formulations. It simulated five LNP formulations, yielding mean apparent pKa values with a 0.5 unit error compared to experiments. This method enhances understanding of ionizable lipids’ charge distributions, impacting drug delivery and manufacturing.

  • Generative Design of Singlet Fission Materials by Revisiting the Use of a Fragment-oriented Database
    Thanapat, Worakul; Rubén, Laplaza; J. Terence, Blaskovits; Clémence, Corminboeuf
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-20
     The paper utilizes the FORMED repository of 116,687 synthesizeable molecules to conduct fragment-oriented virtual screening and genetic algorithms for identifying singlet fission materials. By employing reinforcement learning, the authors rediscover various known chromophores and propose a novel compound, neocoumarin, displaying diradicaloid behavior. The results emphasize the efficacy of generative design in expanding material discovery.

  • Quantifying design principles for light-emitting materials with inverted singlet-triplet energy gaps
    Laura, McCaslin; Varun, Rishi; Ali, Abou Taka; Hrant, Hratchian
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-20
     The paper explores design principles for light-emitting materials with inverted singlet-triplet gaps (ΔEST). Using the HEPTA-INVEST15 dataset of 15 heptazine-based molecules, it finds a strong correlation (R2 > 0.94) between intramolecular charge transfer and ΔEST. Key findings highlight the impact of functional groups on ΔEST, offering predictive metrics for future organic emitter design using machine learning methods.

  • Dual-sites for low-concentration NO to NH3 electrosynthesis in neutral media: promoting NO adsorption and water dissociation
    Tongwei, Wu; Xiaoxi, Guo; Hengfeng , Li; Yanning , Zhang; Chao, Ma; Hongmei , Li; Haitao, Zhao; Liyuan, Chai; Min, Liu
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-20
     The paper presents a novel dual-site catalyst combining Pt nanoparticles (PtNPs) and an Fe single-atom catalyst (FeSAC) to enhance electrosynthesis of ammonia (NH3) from low-concentration NO in neutral media. Methods include in situ X-ray absorption and infrared spectroscopy. Results show a Faradaic efficiency of 90.3% and a record NH3 yield of 3023.8 µg h-1 mgcat.-1, surpassing traditional catalysts.

20 Mar 2025

  • Molecularly defined glycocalyx models reveal AB5 toxins recognize their target glycans superselectively
    W. Bruce Turnbull, Laia Saltor Nf1ez, Vajinder Kumar, James F. Ross, Jonathan P. Dolan, Sumitra Srimasorn, Xiaoli Zhang, Ralf P. Richter
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-19
     The paper discusses the development of glycocalyx models utilizing mucin-like glycopolymers to study AB5 toxins like cholera and shiga toxins. Methods employed include quartz crystal microbalance and spectroscopic ellipsometry to analyze toxin binding kinetics. Results indicate that toxins show significantly enhanced affinity in these models, suggesting superselective binding critical for host cell selection.
  • Evolutionary Machine Learning of Physics-Based Force Fields in High-Dimensional Parameter-Space
    David van der Spoel, Julian Marrades, Kristian Kriz, A. Najla Hosseini, Alfred Nordman, Joao Paulo Ateide Martins, Paul J. van Maaren, Mohammad Mehdi Ghahremanpour, Marie-Madeleine Walz
    Organic Chemistry on ChemRxiv
    2025-03-19
     The paper introduces the Alexandria Chemistry Toolkit (ACT), an open-source software for machine learning physics-based force fields (FFs) using customizable potential functions. It employs a genetic algorithm and Monte-Carlo methods to optimize FF parameters. The evaluation of different models demonstrates improved prediction accuracy for organic molecules’ properties in gas and liquid phases.
  • MASSISTANT: A Deep Learning Model for De Novo Molecular Structure Prediction from EI-MS Spectra via SELFIES Encoding
    Marcin Pietrasik, John Mommers, Lazar Barta, Anna Wilbik
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-19
     The paper presents MASSISTANT, a deep learning model for predicting molecular structures from low-resolution EI-MS spectra using SELFIES encoding. Using the NIST dataset with 180k spectra, MASSISTANT achieves about 10% exact predictions, improving to 54% with a curated dataset. The model enhances the interpretation of chromatographic peaks in mass spectrometry, leveraging complex fragmentation patterns to generate valid chemical structures.
  • Data-driven discovery of water-stable metal-organic frameworks with high water uptake capacity
    Akash Ball, Gianmarco Terrones, Shuwen Yue, Heather Kulik
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-19
     The paper presents a machine learning and high-throughput screening approach to discover water-stable metal-organic frameworks (MOFs) with high water uptake capacity. Utilizing 736 curated MOFs and grand canonical Monte Carlo simulations, the study identifies strong correlations between pore features and water uptake. Ultimately, 74 promising MOFs are predicted to exhibit both water stability and high capacity.
  • 3D2SMILES: Translating Physical Molecular Models into Digital DeepSMILES Notations Using Deep Learning
    Wenqi Guo, Yiyang Du, Mohamed Shehata
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-19
     The paper presents 3D2SMILES, a deep learning model that converts images of physical molecular models into digital DeepSMILES notations. Utilizing both synthetic and real datasets, it achieved 62.0% top-1 and 80.3% top-3 accuracy. The study also addresses model explainability, limitations, and future research directions.

19 Mar 2025

  • Data Scaling and Generalization Insights for Medicinal Chemistry Deep Learning Models
    Alan Cheng, Jacky Chen, Song Yang, Jonathan Tynan, Cheng Chen, Yunsie Chung
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-18
     The paper investigates deep learning models for drug discovery, comparing them to traditional methods like XGBoost and random forest using internal and publicly available datasets. It examines performance across various data tasks and identifies factors boosting graph neural network performance. A scaling relationship accounting for 81% of performance variance is developed, aiding in estimating models for ADMET endpoints and enhancing drug discovery outcomes.
  • Machine Learning-Guided Synthesis of Prospective Organic Molecular Materials: An Algorithm with Latent Variables for Understanding and Predicting Experimentally Unobservable Reactions
    Norie MOMIYAMA, Kazuhiro TAKEDA, Naoya OHTSUKA, Toshiyasu SUZUKI
    Organic Chemistry on ChemRxiv
    2025-03-18
     The paper presents a machine learning algorithm that incorporates latent variables to predict unobservable reactions for perfluoro-iodinated naphthalene derivatives. The algorithm achieved an R2 value >0.99, accurately estimating substitution patterns and reaction yields, validated experimentally. This ML-guided framework enhances heuristic methods in chemistry, optimizing synthetic processes and enabling further applications in catalyst discovery and organic semiconductor optimization.
  • Transferable Machine Learning Interatomic Potential for Pd-Catalyzed Cross-Coupling Reactions
    Olexandr Isayev, Dylan Anstine, Roman Zubatyuk, Liliana Gallegos, Robert Paton, Olaf Wiest, Benjamin Nebgen, Travis Jones, Gabe Gomes, Sergei Tretiak
    Organic Chemistry on ChemRxiv
    2025-03-18
     The paper presents AIMNet2-Pd, a machine learning interatomic potential for palladium-catalyzed cross-coupling reactions, enabling rapid computational studies. Trained on monophosphine ligands, it performs energy calculations and geometry optimizations with high accuracy (1-2 kcal mol−1 and ~0.1 Å) in seconds, facilitating high-throughput catalyst screening and mechanistic studies of diverse Pd complexes.
  • Ascorbate Peroxidase (APEX2) Activates Dihydrotetrazine Oxidation for Rapid Bioorthogonal Chemistry in Living Cells
    Joseph Fox, Sophia Neglia, Amanda Tallon, Christopher am Ende, Colin Thorpe
    Organic Chemistry on ChemRxiv
    2025-03-18
     The paper demonstrates that APEX2 efficiently catalyzes dihydrotetrazine (DHTz) oxidation, enhancing bioorthogonal reactions in living cells. Through in vitro studies, it reveals the role of superoxide dismutase and the impact of H2O2 on reaction efficiency (kcat/KM 4.90 × 10^3 M−1s−1). Tests included live HeLa and PC3 cells, showing successful tetrazine conjugation with fluorescent tags and selective targeting of enzymes.
  • Addressing sustainability challenges in peptide synthesis with flow chemistry and machine learning
    Kristóf Ferentzi, András Perczel, Viktor Farkas
    Organic Chemistry on ChemRxiv
    2025-03-18
     The paper presents a recyclable Anisole/DMSO solvent system for flow chemistry in peptide synthesis, outperforming DMF. Key methods included solvent parameter exploration, flow parameter adjustments reducing racemization, and machine learning optimization. High-efficiency synthesis of peptides (Aib-ACP, GLP-1, BPTI) was achieved, making the process more sustainable by minimizing toxic waste and improving speed.
  • Racemization-free peptide bond formation via 2-nitrobenzensulfonyl strategy for diastereoselective synthesis of (Z)-fluoroalkene-type peptidomimetcs
    Tetsuo NARUMI, Chihiro Iio, Kohei Sato, Nobuyuki Mase
    Organic Chemistry on ChemRxiv
    2025-03-18
     The paper discusses a novel racemization-free strategy for synthesizing (Z)-fluoroalkene-type peptidomimetics, using the 2-nitrobenzensulfonyl (Ns) group to prevent racemization during peptide bond formation. The method involves coupling Xaa-Pro-type FADIs with amino acid benzyl esters, enhancing diastereoselectivity and expanding the applications of FADIs in peptide development.
  • Chiral Potassium Brønsted Base-Catalyzed Stereoselective Synthesis of 1,3-Diols via a Tandem Allylic Isomerization/Asymmetric Aldol–Tishchenko Reaction
    Masahiro Sai, Hiroki Ishikawa
    Organic Chemistry on ChemRxiv
    2025-03-18
     The paper discusses chiral potassium Brønsted bases featuring 3,3́-substituted BINOL-based crown ethers as sustainable catalysts for a tandem allylic isomerization/asymmetric aldol–Tishchenko reaction. This method produces diverse 1,3-diols with excellent diastereoselectivity and high enantioselectivity, utilizing allylic alcohols as nucleophiles.
  • An Electrochemical Amidation/C-H Halogenation Cascade for the Sustainable Synthesis of Halogenated N-Aryl Amides, Carbamates and Ureas
    Oscar Verho, Sudipta Ponra, Ruzal Sitdikov, Alyssio Calis
    Organic Chemistry on ChemRxiv
    2025-03-18
     The paper presents an electrochemical cascade method for synthesizing halogenated N-aryl amides, carbamates, and ureas, combining amide bond formation with C-H chlorination. Demonstrating its robustness, over 180 products were synthesized, highlighting its scalability and environmental benefits compared to traditional methods.

18 Mar 2025

  • Molecular deep learning at the edge of chemical space
    Derek van Tilborg, Luke Rossen, Francesca Grisoni
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-17
     The paper proposes a joint modeling approach to enhance molecular machine learning’s ability to generalize beyond training data. Utilizing over 30 bioactivity datasets, it introduces a reconstruction-based “unfamiliarity” metric to predict model performance on novel bioactive molecules. The results show that this method effectively identifies out-of-distribution molecules and offers deeper molecular insights compared to traditional approaches.
  • How repair proteins identify DNA damage in the nucleosome
    Fabrizio Cleri, Parvathy A. P. Sarma, Safwen Ghediri, Vinnarasi Saravanan, Corinne Abbadie, Ralf Blossey
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-17
     The paper investigates how repair proteins identify DNA damage in nucleosomes using molecular dynamics simulations. It focuses on interactions between glycosylase UDG and mutated uracil, and PARP1 with simulated single-strand breaks. Results suggest that specific mechanical deformations around DNA defects may enhance damage recognition, making random search insufficient for efficient repair initiation.
  • Neural Mulliken Analysis: Molecular Graphs from Density Matrices for QSPR on Raw Quantum-Chemical Data
    Oleg Gromov
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-17
     The paper introduces Neural Mulliken Analysis, leveraging molecular graphs from one-electron density matrices for QSPR. Utilizing atomic and link node embeddings, the method employs a GNN trained on data from the Solubility Challenge, achieving improved solubility predictions (RMSE 0.63, R^2 0.79). This approach shows potential for enhancing chemical machine-learning tasks by integrating electronic structure awareness.
  • Leveraging Alchemical Free Energy Calculations with Accurate Protein Structure Prediction
    Jan Domański, Jenke Scheen, Francesco Rianjongdee, Hannah Bruce Macdonald, Richard Gowers, Seb Degorce, Adam Green, Conor Scully, Tom Duffy, Jen Howes, Charlotte Cordery, Alwin Bucher, Laksh Aithani
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-17
     The paper presents a hybrid framework combining machine-learned ligand-protein co-folding models with Free Energy Perturbation (FEP) for drug discovery. It benchmarks results on a public kinase target (PFKFB3) and an internal target, showing improved accuracy and reduced computational costs compared to traditional methods, thus enhancing lead optimization and supporting therapeutic discovery.
  • Time-Series Learning Based on Neural Ordinary Differential Equations for Nonadiabatic Molecular Dynamics Simulations
    Haoyang Xu, Jin Wen, Luxiang Zhu, Lei Niu, Feng Yan, Meifang Zhu
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-17
     The paper presents two time-series learning models based on neural ordinary differential equations (NODEs) for nonadiabatic molecular dynamics simulations: continuous normalizing flow (CNF) and Hamiltonian neural networks (HNN). CNF improves modeling of energy gap distributions, while HNN effectively handles nonadiabatic transitions. Both architectures show potential for ultrafast dynamics simulations, with CNF simplifying parameter tuning and HNN enhancing configurational reorganization simulations.

14 Mar 2025

  • mineMS2: Annotation of spectral libraries with exact fragmentation patterns
    Etienne A., Thévenot, Alexis, Delabrière, Coline, Gianfrotta, Sylvain, Dechaumet, Pierrick, Roger, Annelaure, Damont, Thaïs, Hautbergue, Emilien L., Jamin, Olivier, Puel, Christophe, Junot, François, Fenaille
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-13
     The paper presents mineMS2, a software for annotating spectral libraries using exact fragmentation patterns. It employs a novel graph-based representation and a frequent-subgraph mining algorithm. Evaluated on spectral databases and biological samples, mineMS2 identifies similarities overlooked by traditional methods, enhancing structural interpretation and the identification of unknown metabolites. The software is available as an R package.

  • HyperXtract: Strategic Platform for Optimizing High-Bandwidth Nanopore Data Extraction Performance
    Y.M. Nuwan, D.Y. Bandara, Katherine, Gussenhoven, Dhanush, Amarasekara, Jugal, Saharia
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-13
     The paper introduces HyperXtract (HX), a platform optimizing high-bandwidth nanopore data extraction, outperforming its predecessor, NanoPlex (NP). HX processes 1000-second files at 40 MHz in under 50 seconds, significantly faster than NP’s >45× slower performance. Key methods include MEX functions, memory mapping, and pre-allocation for efficient analysis, allowing near real-time decision-making in high-bandwidth experiments.

  • Graph neural networks to predict atomic transition charges and exciton couplings in organic semiconductors
    Daniel, Packwood, Geoffrey, Weal, Maryam, Nurhuda, Justin, Hodgkiss, Paul, Hume
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-13
     The paper presents a graph neural network (GNN) model that predicts exciton couplings in organic semiconductors using atomic transition charges. It demonstrates successful predictions for fused-ring electron acceptors and other molecules from the Cambridge Crystallographic Data Centre, enabling accurate exciton diffusion simulations and advancing high-throughput virtual screening for photovoltaic applications.

  • Comparison of Magnesium and Manganese Ions on the Structural and Catalytic Properties of Human DNA Polymerase Gamma
    Arkanil, Roy, G. Andrés, Cisneros
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-13
     The paper compares the effects of magnesium (Mg²⁺) and manganese (Mn²⁺) on human DNA polymerase gamma (Pol γ) using molecular dynamics and quantum mechanics/molecular mechanics simulations. Mn²⁺ enhances catalytic efficiency and flexibility but reduces structural stability compared to Mg²⁺. Findings indicate a trade-off between these properties, with implications for mutagenesis and mitochondrial disorders.

  • Cooperative Free Energy: Induced Protein–Protein Interactions and Cooperative Solvation in Ternary Complexes
    Sereina, Riniker, Shu-Yu, Chen, Riccardo, Solazzo, Marianne, Fouché, Hans-Jörg, Roth, Birger, Dittrich
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-13
     The paper investigates cooperative free energy in ternary complexes, emphasizing induced protein-protein interactions (PPIs) and cooperative solvation. Using an approximated expression, it developed a rapid computational method to predict cooperativity in eight complexes, achieving a Kendall τ of 0.79. The study critiques current interpretations of cooperativity and provides thermodynamic insights into protein-ligand-protein interactions.

13 Mar 2025

  • Conformal Selection for Efficient and Accurate Compound Screening in Drug Discovery
    Tian, Bai; Peng, Tang; Yuting, Xu; Vladimir, Svetnik; Abbas, Khalili; Xiang, Yu; Archer, Yang; Bingjia, Yang
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-12
     The paper presents a conformal selection approach for compound screening in drug discovery, addressing bias and risk control issues. By using conformal inference to generate p-values, it ensures rigorous false discovery rate control and high accuracy. The method shows improved computational efficiency and is validated through numerical simulations on real-world datasets.
  • DeepDeg: Forecasting and explaining degradation in novel photovoltaics
    Felipe, Oviedo; David S., Hayden; Thomas, Heumeuller; Jonas, Wortmann; Jose Dario, Perea; Richa, Naik; Hansong, Xue; John, Fisher III; Christoph J, Brabec; Tonio, Buonassisi; Juan, Lavista
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-12
     The paper presents DeepDeg, a machine learning model that forecasts and explains degradation in novel photovoltaics. Using a dataset of over 785 stability tests (230,000 hours), it accurately predicts degradation dynamics and explains the factors involved. Evaluated with 9,000 hours of data, DeepDeg accelerates degradation characterization by 5-20 times.
  • Automated Predictive Chemical Reaction Modelling applied to Gold(I) Catalysis
    Raphaël, Robidas; Claude Y., Legault
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-12
     The paper presents a theoretical framework for automated predictive chemical reaction modeling applied to gold(I) catalysis, utilizing a “neophile” kinetic model. It achieved successful modeling of 17 unimolecular reactions, closely matching experimental results in 11 cases. This approach aims to facilitate unbiased reaction data generation and future machine-learning applications for autonomous research.
  • Exploring Chemistry and Catalysis by Biasing Skewed Distributions via Deep Learning
    Zhikun, Zhang; GiovanniMaria, Piccini
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-12
     The paper presents Loxodynamics, a deep learning approach for automating chemical reaction discovery using biased molecular dynamics. It utilizes a skewed distribution to identify low-energy reaction pathways, employing Skewencoder for efficient data extraction. Validated on various reactions, including gas-phase and catalytic processes, Loxodynamics provides a novel, data-driven framework for navigating complex chemical systems.
  • Harnessing Electric Fields for Rare Earth Element Recovery: A Computational Study of Electromigration and Flow Dynamics Using Dilute Feedstock
    Venkateshkumar, Prabhakaran; Zirui, Mao; Zhijie, Xu; Yang, Huang; Giovanna, Ricchiuti; Bruce, Palmer; Pauline G. , Simonnin; Vijayakumar, Murugesan; Grant E, Johnson; Jaehun , Chun
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-12
     The paper presents a Computational Fluid Dynamics (CFD) study exploring electrohydrodynamic methods for separating rare earth elements (REEs) from dilute feedstocks. It identifies key separation regimes influenced by diffusion, advection, and electromigration. Optimal conditions for separation require slow flows and strong electric fields. Results emphasize the significance of channel design and flow dynamics for effective REE recovery.
  • Comparing Massively-Multitask Regression Algorithms for Drug Discovery
    Eric, Martin; Xiang-Wei, Zhu; Patrick, Riley; Steven, Kearnes; Ekaterina A , Sosnina; Li, Tian; Ying, Wei; Thomas M, Whitehead; Gareth J, Conduit; Matthew D, Segall
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-12
     The paper compares six Massively-Multitask Regression Models (MMRMs) for bioactivity prediction in drug discovery, using datasets of 159 kinase and 4276 diverse ChEMBL assays. Results show MMRMs outperform single-task random forests, particularly in hit-finding, but not for virtual screening. Training on 99+% of data yields better accuracy than training on 75%. Practical recommendations for method selection are discussed.

12 Mar 2025

  • Validation of dried saliva for molecular diagnostics
    Charles R., Mace; Andrea C., Mora; Alexandra K., Sogn; Allison J., Tierney; Elizabeth, Tzavaras; Mabi L., Singh; Gustavo, Mahn Arteaga; Fiorenzo G., Omenetto; Athena, Papas
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-11
     The paper discusses the validation of FishburneTabs for standardized saliva sampling in viral diagnostics, particularly for COVID-19. It involves a method where patients collect saliva on paper tabs that dry and can be processed like dried blood spots. With 125 clinical samples, this method achieved 85% sensitivity and 94% specificity, offering a reliable alternative to liquid saliva for decentralized testing.

  • A one-pot RT-LAMP diagnostic assay for SARS-CoV-2 from saliva samples
    Charles R., Mace; Andrea C., Mora; Allison J., Tierney; Alexandra K., Sogn; Paul T., Lawrence; Elizabeth, Tzavaras; Mabi L., Singh; Gustavo, Mahn Arteaga; Fiorenzo G., Omenetto; Athena, Papas
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-11
     The paper describes a one-pot RT-LAMP assay for detecting SARS-CoV-2 in saliva, addressing challenges of saliva’s complexity. The method involves five steps, including heating and reaction processes. Tested on 127 patient samples, the assay achieved 98% accuracy, 88% sensitivity, and 100% specificity, demonstrating potential for improved point-of-care diagnostics for respiratory viruses.

  • Novel Tacrine-Based Compounds: Bridging Cholinesterase Inhibition and NMDAR Antagonism
    Jan, Korabecny; Barbora, Svobodova; Zuzana, Moravcova; Anna, Misiachna; Gabriela, Novakova; Ales, Marek; Vladimir, Finger; Jitka, Odvarkova; Jaroslav, Pejchal; Jana, Karasova Zdarova; Jakub, Netolicky; Marek, Ladislav; Martina, Hrabinova; Ales, Sorf; Lubica, Muckova; Lenka, Fikejzlova; Marketa, Benkova; Martin, Novak; Lukas, Prchal; Jan, Capek; Jiri, Handl; Tomas, Rousar; Katarzyna, Ewa Greber; Krzesimir, Ciura; Martin, Horak; Ondrej, Soukup
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-11
     The paper discusses the design and evaluation of 16 novel tacrine derivatives for Alzheimer’s disease treatment, focusing on combining cholinesterase inhibition and NMDAR antagonism. In silico screening and in vitro assays assessed bioavailability and safety, revealing compound 5m as a potent inhibitor with improved metabolic stability and CNS penetration. The study supports further development of these multi-target ligands.

  • DNA-programmable Protein Degradation: Dynamic Control of PROTAC Activity via DNA Hybridization and Strand Displacement
    Michael, Booth; Disha, Kashyap; Thomas, Milne
    Organic Chemistry on ChemRxiv
    2025-03-11
     The paper presents OligoPROTACs, DNA-linked PROTACs that enable controlled protein degradation through DNA hybridization and a dynamic off-switch mechanism via strand displacement. The study showcases how these constructs can achieve distance-dependent degradation and reduce off-target effects, suggesting DNA nanotechnology’s potential to refine PROTAC systems for safer therapeutic applications.

  • Enzymatic Crosslinking of Polyelectrolyte Complexes Produces Strong, Reconstitutable Biomedical Adhesives
    Aasheesh, Srivastava; Tanmay , Dutta; Bhanwar, Lal; Drishya, Vats; Roop Singh, Lodhi; Paramita, Das; Debasis, Nayak
    Biological and Medicinal Chemistry on ChemRxiv
    2025-03-11
     The paper presents a novel water-based biomedical adhesive formed from anionic and cationic polyelectrolytes with Horseradish Peroxidase. It demonstrated strong adhesion (up to 10 MPa) on various surfaces and superior hemostatic properties in vivo compared to commercial adhesives. The adhesive is also reconstitutable from a lyophilized form, indicating potential for effective wound healing applications.

11 Mar 2025

  • Machine Learning-Based Retention Time Prediction Tool for Routine LC-MS Data Analysis
    Oleksandr O., Grygorenko; Sofiia A., Dymura; Oleksandr O., Viniichuk; Kostiantyn P., Melnykov; Dmytro S., Radchenko
    Organic Chemistry on ChemRxiv
    2025-03-10
     The paper presents a machine learning model for predicting retention time (RT) in LC-MS data analysis using a GATv2Conv+DL graph neural network. Trained on internal data and evaluated with the METLIN SMRT dataset, the model achieved a mean absolute error of 2.83 seconds. It was integrated into an existing LC-MS tool, with 95% of data within the RT range of -11.34 to +10.68 seconds.

  • Investigating Cavity Quantum Electrodynamics-Enabled Endo/Exo- Selectivities in a Diels-Alder Reaction
    Pengfei, Huo; Jialong, Wang; Braden, Weight
    https://chemrxiv.org/engage/rss/chemrxiv?categoryId=605c72ef153207001f6470ce
    2025-03-10
     The paper explores how cavity quantum electrodynamics can alter selectivity in a Diels-Alder reaction, using a parameterized quantum electrodynamic (pQED) approach. It shows that coupling to the cavity changes selectivity from equal probabilities of Endo/Exo isomers outside the cavity. The findings align with high-level QED methods, demonstrating electron density redistribution due to the cavity effects.

  • When Theory Came First: A Review of Theoretical Chemical Predictions Ahead of Experiments
    Mario, Barbatti
    https://chemrxiv.org/engage/rss/chemrxiv?categoryId=605c72ef153207001f6470ce
    2025-03-10
     The paper reviews twenty examples over fifteen years where theoretical predictions in computational chemistry anticipated experimental findings in fields like bioinorganic chemistry and materials science. It highlights the essential role of quantum chemical methods in predicting molecular structures, reaction mechanisms, and material properties, demonstrating their impact on guiding experiments and scientific discovery.

  • On the Reproducibility of Free Energy Surfaces in Machine-Learned Collective Variable Spaces
    Matteo, Salvalaglio; Florian M., Dietrich
    https://chemrxiv.org/engage/rss/chemrxiv?categoryId=605c72ef153207001f6470ce
    2025-03-10
     The paper discusses reproducibility in machine-learned collective variable (MLCV) spaces for calculating free energy surfaces (FES). It identifies the influence of training variability and hyperparameter roughness on reproducibility, proposing a Geometric Free Energy representation and a normalization factor to unify definitions and improve gradient modeling. These approaches aim to enhance the adoption of MLCVs in molecular simulations.

07 Mar 2025

  • Evaluating Machine Learning Models for Molecular Property Prediction: Performance and Robustness on Out-of-Distribution Data
    Johannes, Kirchmair; Hosein, Fooladi; Thi Ngoc Lan, Vu
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-06
     The paper evaluates twelve machine learning models, including random forests and graph neural networks, on eight datasets to assess their robustness on out-of-distribution (OOD) data. It reveals that OOD data, particularly chemical similarity clustering, poses challenges for model performance. The correlation between in-distribution (ID) and OOD performance varies significantly, informing model selection based on application domains.

  • Enhancing PROTAC Ternary Complex Prediction with Ligand Information in AlphaFold 3
    Rocío, Mercado; Francisco, Erazo; Nils, Dunlop; Farzaneh, Jalalypour
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-06
     The paper discusses enhancing AlphaFold 3’s predictions of PROTAC ternary complexes by incorporating ligand information. Evaluating 48 PROTAC-related structures from the Protein Data Bank, the authors demonstrate improved accuracy using metrics like RMSD and pTM when ligand details are included. They also assess input strategies for better predictions, releasing their methodology and results on GitHub.
  • RIGR: Resonance Invariant Graph Representation for Molecular Property Prediction
    Akshat Shirish, Zalte; Hao-Wei, Pang; Anna C, Doner; William H., Green
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-06
     The paper introduces RIGR, a Resonance Invariant Graph Representation that improves molecular property predictions by creating a single representation for resonance-exhibiting molecules. Evaluated on a large dataset using the D-MPNN architecture, RIGR outperforms Chemprop while using 60% fewer features. It is available as an open-source featurization scheme, showing promise in various property prediction tasks.

06 Mar 2025

02 Mar 2025

  • Evaluating Machine Learning Models for Molecular Property Prediction: Performance and Robustness on Out-of-Distribution Data
    Johannes Kirchmair, Hosein Fooladi, Thi Ngoc Lan Vu
    Theoretical and Computational Chemistry on ChemRxiv
    2025-03-01
     This study evaluates twelve machine learning models for molecular property prediction, focusing on their performance on out-of-distribution (OOD) data using eight datasets and seven splitting strategies. It finds that while classical and GNN models perform similarly under scaffold-based splitting, cluster-based splitting presents more challenges. The correlation between in-distribution and OOD performance varies significantly, emphasizing careful model selection for specific applications.
Search