This Week In Cheminformatics: Issue #033

Andreas Bender and others argue in this perspective that AI has meaningful use cases in drug discovery without _yet_ showing evidence of clinical impact, and why that might be. The paper builds on previous works of Andreas (read his blogpost about them) including the cost model that suggests optimizing phase II success rates is the most fruitful strategy (monetarily). Most of the “AI” work is in preclinical hit finding, i.e., in ligand discovery rather than drug discovery, where labelled data happens to exist. The paper also makes the data case i.e. biological data is conditional on assay parameters, cell system, dose and patient context, which makes label assignment unreliable or at best uncertain; chemical data is sparse and biased, for instance, the four common ADME datasets from Therapeutics Data Commons share only 0.1% of compounds, so applicability domains might not line up across endpoints in a multi-objective setting. The paper most interestingly argues model validation is not process validation and that’s a really neat way to think about the problem. Great read as always.

Some related links:

• Read the paper here: https://rdcu.be/fyr77

Eric Topol@EricTopol

A in-depth review of AI for drug discovery "It can be concluded that we are still at a stage with an ‘absence of evidence’ (and not necessarily an 'evidence of absence') when it comes to the translation of AI into clinically relevant impact." @NatRevDrugDisc

3:46 PM · Aug 9, 2026 · 29.3K Views

19 Replies · 56 Reposts · 243 Likes

• So How Is AI Drug Discovery Doing, Really? by Derek Lowe

• Thibault Geoui’s video on LinkedIn

• Andreas’s LinkedIn post

Armstrong, Nguyen et al. from Schwaller Group present SynthEx as a retrosynthesis planner built from LLM agents rather than a template library. The model writes each disconnection as a list of atom-level graph edits, and those edits are applied to the product to produce the precursors. The authors ran it on 1,098 natural products from NP-Atlas that have no published total synthesis. AiZynthFinder, finds complete routes for ~14% of them. SynthEx finds routes for ~64%. Ten synthetic chemists rated key steps blind, mixed in with steps from total syntheses published after the model’s cutoff, and could not tell them apart on feasibility / overall quality. None of it has been run in a lab, yet. The routes are public as SynthAtlas. Interesting read.

CheMeleon has a new citation now :) If you haven’t used it yet, give it a go here: https://github.com/JacksonBurns/chemeleon and https://chemprop.readthedocs.io/en/latest/chemeleon_foundation_finetuning.html . Good read!

• Generative virtual screening: from search to generate and back — Great read !!

• Multi-task graph neural networks for comprehensive surfactant property prediction

• A scalable open-source software platform for managing and automating massively parallel battery cycling experiments

• A decoupled alignment kernel for peptide membrane permeability predictions

• Train–test splits matter more for evaluation than for performance

• ALF: Open-Source Active Learning Framework for Atomistic Modeling

• Explicit, Machine-Learned Two-Body Potentials for Molecular Simulations

• One-Dimensional Molecular Coordinate Orders Amino Acids across

  Water-Mediated Processes

• ProtLID2.0: A Residue-Specific Pharmacophore Framework for Protein–Protein Docking

• An OpenMM-Based ML/MM–MMGBSA Workflow for End-point Protein–Ligand Binding Energy Ranking

• Motif-Based Graph Learning for Synthetic Reaction Condition Prediction

• Deep3MVPF: Multiview Deep Framework for the Prediction of Stability and m6A in mRNA 3′UTR

• Odor-Sight: An Interpretable Graph Neural Network Web Platform

  for Prediction of Odorant Activity

• Uncertainty Quantification with Domain Classification Models for Acetylcholinesterase Inhibition

• MK-ensemble: fragment-based multi-kernel ensemble for interpretable structure-activity relationship modeling of steroidal saponins

• AI-guided De Novo design of DHPS inhibitors

• Novelty Assessment Method for Streaming Samples Based on Slope Entropy

• Geometric Deep Learning-Based Drug Design Models for Small-Molecule Drug Discovery

• PairMap2: A Web Application for Intermediate-Molecule Insertion in Relative Binding Free Energy Calculations

• pyMSscreen: an interactive interface for automated quality control and prescreening in mass spectrometry workflows

• Development of a Conformational Sampling Method for Transition State Structures With Diverse Active Bond Lengths

• Phyfold: environment aware mathematical modeling for protein folding dynamics integrated with physics informed neural network

• Machine learning–guided QSAR screening of fluconazole analogs and FDA-approved drugs against Candida albicans, with docking, molecular dynamics, and ADMET analysis

• Fast intermolecular interaction energy calculation with the OPLS-AA force field

• Drug target prediction from perturbation transcriptomics via a biological function-guided hypergraph siamese network

• Teeter-Totter-Like Redistribution of Switch I: Integrating Simulation and Experiment to Uncover the Molecular Basis of Oncogenic Rac1Mutations

• Integrating AI and Molecular Modeling for Structural Prediction of a

  Closed-Pore State of the hERG Channel

• Buffer Region Embedding for Hybrid Machine-Learned/MolecularMechanical Simulations in Complex Environments

• IAMV-DDI: Interaction-Aware Multi-View Molecular Representation Learning for Drug–Drug Interaction Event Prediction

• Unraveling the Molecular Basis of NSD3 Allosteric Activation Driven by Nucleosome Binding: A Multiscale MD and QM/MM Approach

• Computational Insights into the Dissociation Pathway of the Charybdotoxin–Shaker Complex: Voltage Dependency and Potassium Modulation

• TriGraphQA: A Triple Graph Learning Framework for Model Quality Assessment of Protein Complexes

• Fragment-Based Discovery of KLK6 and KLK7 Inhibitors

• Exploring Competitive Protein–Protein Interaction Mechanisms through Dynamic Residue Energy Landscapes for Antizyme Design and Validation

• Bend and Snap: Computationally Exploring the Helix Bias for Classifying Abl Kinase Allosteric Modulators

• CBInformax: bioactivity-aware self-supervised molecular representation learning for molecular property and drug–drug interaction prediction

• Reliable Identification and Out-of-Library Detection in Mass Spectra

• Explicitness in SMILES representation via ExACT: improved tokenization for aqueous solubility prediction

• Targeting the RNA-Binding Site of SARS-CoV‑2 NSP13 by FRASE-bot

  in CACHE Challenge #2

• Barrel Shape and Chromophore Rigidity Predict Fluorescent-Protein

  Photophysics

• Computational Study of Allosteric Switch of NLRP3: How ATP Binding Unlocks the NACHT Domain

• Surface Adsorption and Geometry Evaluator (SAGE): An Interactive Workflow Platform for Standardized Surface Adsorption Screening

• MVGCL: Noise-Robust Multi-Source Similarity Fusion and Type-Aware Dual-Pathway Learning for Drug Repositioning

• Harmonizing Peak Matching between Multidimensional NMR

  Spectra

• NextTopDocker: A Large-Scale Docking-Power Benchmark Reveals Limitations of Current End-to-End Machine-Learning Docking and the Strength of Hybrid Rescoring

• PepGate: A Dual-Path Diffusion Framework for ACE Inhibitory Peptide De Novo Design

• Neutral Backbone Modifications Enhance the Pharmacokinetics and

  Biodistribution of Antibody−Oligonucleotide Conjugates with High

  Drug-to-Antibody Ratios

• From Porphyrins to Versatile Photostable Bacteriochlorins: TripletState-Driven Design of Perfluorinated Photosensitizers

• Deciphering Chemical Environments in Conformational Data Sets Using an IQA Band-Matching Protocol

• Tuning Cyclopentadiene Isomerization by Substituent Effects

• Total Synthesis of (+)-CC-1065

• Harnessing Host–Guest Conformational Matching for Efficient Near-Infrared Phosphorescence

• An Explicit Interaction-Prompted Diffusion Framework for High-Fidelity 3D Molecular Generation

François Fleuret@francoisfleuret

[From what I experience] even with current sota-super-smart models, brainstorming about an actually new problem consists for them to bring and try to combine existing bits from the literature, and not to do the real research process. 1/2

8:15 AM · Aug 10, 2026 · 27.1K Views

15 Replies · 1 Repost · 124 Likes

Kevin Lacker@lacker

A mathematician is giving a lecture and says, "It is obvious that…." Then he stops, stares at the board, and thinks silently for ten minutes. Finally he brightens up and says, “Yes! It is obvious.”

8:25 PM · Aug 10, 2026 · 112K Views

9 Replies · 113 Reposts · 5.75K Likes

Programmer Humor@PR0GRAMMERHUM0R

slowClap

12:00 AM · Aug 10, 2026 · 19.1K Views

32 Reposts · 994 Likes

Adam Karpiak@Adam_Karpiak

the internal candidate waiting for you to finish your 6th interview

2:04 AM · Aug 10, 2026 · 1.82M Views

61 Replies · 3.49K Reposts · 79.8K Likes

Strace@straceX

we invented virtual memory so programs could pretend RAM is bigger then invented swap so the lie could continue

11:27 PM · Aug 9, 2026 · 71.5K Views

36 Replies · 100 Reposts · 3.3K Likes

Fermat's Library@fermatslibrary

Wolfgang Paul was born exactly 113 years ago today. He invented the quadrupole ion trap ("Paul trap") and won the Nobel Prize in Physics in 1989. On meeting Wolfgang Pauli in Bonn he reportedly quipped "Finally, I meet my imaginary part!"

2:33 PM · Aug 10, 2026 · 63.6K Views

17 Replies · 152 Reposts · 1.64K Likes

Andy Xu ✈️ ICML@4ndyXu

no way I can't pay for a ticket on the amtrack site bc my last name is too short (3 letter minimum 😭)

7:42 AM · Aug 11, 2026 · 2.91M Views

280 Replies · 576 Reposts · 58.9K Likes

astronikos@formrpessimist

Mark Zuckerberg @finkd

I believe everyone should have access to superintelligence, and I wrote a long piece about Meta's philosophy and values for building a positive future for everyone. https://t.co/2ZoNZXZ39T

2:30 AM · Aug 11, 2026 · 42.5K Views

4 Replies · 25 Reposts · 1K Likes

maria@maria_rcks

Spotify? the engineering company?

Spotify Engineering @SpotifyEng

We just launched Xirp, a vendor-neutral agentic development environment. One place to manage agent sessions across @ClaudeDevs, @GeminiApp CLI, and @OpenAI Codex. 1,300+ @Spotify engineers already use it. Now it's available for you to try. Learn more at https://t.co/Hwo8Qqx4OI.

4:13 PM · Aug 10, 2026 · 537K Views

47 Replies · 179 Reposts · 7.3K Likes

Best,
Manas