Chemoinformatics papers

Best of both worlds: An expansion of the state of the art pKa model with data from three industrial partners Publication describes the biggest (and the most carefully curated) dataset of pKa data that was gathered by industrial collaboration from Hoffmann-La Roche, Genentech, and Bayer CropScience (70k values!). They developed corresponding model and carefully curated it. It offers new insights into chemical spaces covered by compounds from all participating companies and public domain, as well as new insights into their influence on model's predictive accuracy. https://onlinelibrary.wiley.com/doi/10.1002/minf.202400088

I think this could be a very useful tool: automated absolute binding free energy workflow using automated alchemistry engine implemented in the A3Fe python package. It uses GROMACS 2021.3 through BioSimSpace (version 2023.0.0) and SOMD (Sire/ OpenMM Molecular Dynamics) for individual MD simulations. GitHub: https://github.com/michellab/a3fe https://chemrxiv.org/engage/chemrxiv/article-details/6670b524c9c6a5c07aafa972

I think this is the best way to give the equations - to show all the notations as a scheme and not as text. You start to understand it 50% faster and do not have to search the whole document where the notation is defined. Took from an article on Sparrow - a new framework for ranking candidates based on synthesis complexity and cost from Coley's group. Nature | Arxiv

AstraZeneca release Pichemist - tool that calculates the isoelectric point (pI) of a peptide/protein using the Henderson-Hasselbalch equation with the predicted pKa values of its amino acids. #SBDD #docking #QSAR https://pypi.org/project/pichemist/0.1.0/

Nice review of prediction of bioactivity prediction in low-data regime. It is deep-learning centric, I believe there were some historical solutions, but still quite helpful. #deeplearning #QSAR #smalldata #fewshotlearning #LBDD https://www.sciencedirect.com/science/article/pii/S0959440X24000459

Don't forget that separate channel is used for job positions. Please forward me found opportunities, I'm happy to share.

By some reason I often run into job/position/postdoc/internship offers in chemoinformatics. So, I decided to keep everything in one place and not to repost everything in my LInkedIn, which may look weird to my colleagues and friends 😊 I know that somebody here look for a position, so please feel free to join. Also, feel free to add the positions you found. Notice that I don't have much time to keep things in order, so I plan mostly to copy-paste https://t.me/chemoinfo_jobs

Nice publication about binchmarking different single step retrosynthesis approaches. The authors introduced some metrics and benchmarking criteria, and compared major template-based, template-free and semi-template approaches. Also they applied model interpretation approaches to understand why certain predictions were made. "The evaluation on the USPTO-50k case study revealed that frameworks using reaction knowledge, especially in the form of templates, demonstrate superior performance, yielding chemically viable and diverse predictions. Conversely, frameworks relying on deep learning architectures like Transformers and Graph Neural Networks encounter challenges in predicting feasible molecules and reactions. An investigation of model interpretability highlighted the limitations of Transformer models in understanding specific functional groups, possibly limiting their ability to propose novel reactions. Graph-based models performed better in recognising critical motifs, contributing to product stabilisation – but faced challenges with complex reactions." #reactioninformatics #CASD #retrosynthesis #synthesisplanning https://pubs.rsc.org/en/content/articlelanding/2024/dd/d4dd00007b ""

Interesting blog post about alternatives to AlphaFold3. Not sure we can call them all as a real alternative. However, it contains quite good and comprehensive review of approaches in restoring 3D structure of ligand-target complexes. https://medium.com/@falk_hoffmann/promising-alternatives-to-alphafold-3-how-they-work-and-when-to-use-them-ccbcde490b14

Tox24 Challenge announced! The Tox24 challenge is designed to assess the progress in computational methods for predicting in vitro activity of compounds. All ML experts are strongly invited to participate in it and compete for a prize of 1000€, to be awarded for the winning model. #QSAR #drugdesign #data https://ochem.eu/static/challenge.do