Pretrained Molecular Embeddings Offer New Approach to Drug Discovery Similarity Measurement
A team of researchers has introduced pretrained embedding distance (PED), a method for measuring molecular similarity derived directly from pretrained molecular models without task-specific training, accepted at ICML 2026's AI4Science workshop. Traditional similarity measures like Tanimoto fingerprints and 3D shape overlays are either computationally expensive at scale or depend on hand-crafted descriptors, while many deep learning alternatives require costly supervised data curation. PED offers a more general and scalable alternative that could broaden the applicability of AI-assisted drug discovery across diverse biological targets.
Researchers have proposed pretrained embedding distance (PED) as a new approach to molecular similarity measurement in ligand-based drug discovery, with the work accepted at the ICML 2026 AI4Science workshop. Molecular similarity is foundational to tasks such as virtual screening, analog searching, and goal-directed molecular generation, but existing methods face significant trade-offs: fingerprint-based Tanimoto coefficients and 3D shape overlays can be computationally prohibitive at scale, while deep learning methods often require similarity-specific supervision or expensive data curation. PED sidesteps these limitations by computing similarity directly from the embedding space of pretrained molecular models, requiring no additional task-specific fine-tuning. Experimental results demonstrate that PED shows distinct correlations with traditional similarity metrics and performs competitively in both ranking molecules for virtual screening and guiding molecular generation through reward-based design. The authors argue that pretrained molecular embeddings inherently capture rich structural information, making PED a promising and broadly applicable similarity measure. Code and data have been made publicly available alongside the paper.
What's missing
It is unclear how PED compares to traditional methods on computational cost benchmarks at realistic drug-discovery scales, and whether the virtual screening evaluations span a diverse enough set of biological targets to support broad generalizability claims.
What different sources said
- arXiv cs.LGCenter
Advancing Ligand-based Virtual Screening and Molecular Generation with Pretrained Molecular Embedding Distance
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.