Coarse composition suffices: tabular in-context learning for multi-activity antimicrobial peptide profiling
Abstract
Antimicrobial peptides (AMPs) often act against multiple pathogen classes, making multi-label activity prediction a more realistic screening target than binary antimicrobial classification. The ESCAPE benchmark formalizes this setting, but leading approaches typically rely on multimodal, structure-conditioned deep models that are costly to train and tune. We show that a simple, sequence-only pipeline can match and surpass these methods by combining 330 interpretable sequence descriptors with TabPFN, a tabular foundation model that performs in-context prediction in a single forward pass without gradient-based training or hyperparameter search. On ESCAPE (82,359 peptides; five labels), a label-powerset TabPFN model achieves mAP-5 = 77.8%, improving on the previously best reported 72.1%. A probabilistic classifier chain is the first method to match or exceed the best published average precision on each of the five labels simultaneously. The gains persist under the prior state-of-the-art single-fold training protocol, indicating they are not a training-set-size artefact, and are largest for remote homologues (+11.2 points below 30% sequence identity). Ablations further show that predicted structure is unnecessary at inference and that performance is not driven by any single descriptor family: ten global physicochemical scalars recover 91% of full-feature performance. Finally, explicitly modelling label dependence yields targeted benefits for scarce activities and supports ranking which activity to assay next from partial positive evidence.
Related
- Polarity-Guided Extraction of Crithmum maritimum L.: Phenolic Profiling, Selective Anticancer Activity in AR42J and HT-29 Cells, and Antimicrobial Effects
- Cocoa Pulp Juice as a Fermentation Substrate for Potential Probiotic Lacticaseibacillus paracasei WU2502: Genomic Characterization, Metabolite Profiling, and Antimicrobial Activity Against Foodborne Pathogens
- Cell-resolved pharmacodynamic profiling reveals tissue- and cell-population-specific activity of antisense oligonucleotide chemical conjugates in vivo
- Enterococcus faecalis biofilm rewires neutrophil metabolism to suppress antimicrobial activity
- Cell-composition sensitivity of pooled whole-blood DNA methylation profiling in sepsis using the Infinium MethylationEPIC v2.0 array and SeSAMe
BibTeX
@misc{pal2026coarse,
title = {{Coarse composition suffices: tabular in-context learning for multi-activity antimicrobial peptide profiling}},
author = {Anuj Pal and Raunak Kumar and Dhruvi Solanki and Parikshit Pareek and Juhi Singh and Jitin Singla},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2608.30337},
eprint = {2608.30337},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2608.30337},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Coarse composition suffices: tabular in-context learning for multi-activity antimicrobial peptide profiling
AU - Anuj Pal
AU - Raunak Kumar
AU - Dhruvi Solanki
AU - Parikshit Pareek
AU - Juhi Singh
AU - Jitin Singla
PY - 2026
DA - 2026/09/06
PB - arXiv
DO - 10.48550/arxiv.2608.30337
UR - https://arxiv.org/abs/2608.30337
AB - Antimicrobial peptides (AMPs) often act against multiple pathogen classes, making multi-label activity prediction a more realistic screening target than binary antimicrobial classification. The ESCAPE benchmark formalizes this setting, but leading approaches typically rely on multimodal, structure-conditioned deep models that are costly to train and tune. We show that a simple, sequence-only pipeline can match and surpass these methods by combining 330 interpretable sequence descriptors with TabPFN, a tabular foundation model that performs in-context prediction in a single forward pass without gradient-based training or hyperparameter search. On ESCAPE (82,359 peptides; five labels), a label-powerset TabPFN model achieves mAP-5 = 77.8%, improving on the previously best reported 72.1%. A probabilistic classifier chain is the first method to match or exceed the best published average precision on each of the five labels simultaneously. The gains persist under the prior state-of-the-art single-fold training protocol, indicating they are not a training-set-size artefact, and are largest for remote homologues (+11.2 points below 30% sequence identity). Ablations further show that predicted structure is unnecessary at inference and that performance is not driven by any single descriptor family: ten global physicochemical scalars recover 91% of full-feature performance. Finally, explicitly modelling label dependence yields targeted benefits for scarce activities and supports ranking which activity to assay next from partial positive evidence.
ER -