Ontology-grounded nanotoxicity data federation with a provenance-aware knowledge graph and conformal prediction for antimicrobial hazard screening

Nanomaterials are increasingly used across industrial, therapeutic, and consumer applications, while their interactions with biological and environmental systems can produce adverse effects. Although nanotoxicological evidence is abundant, it remains fragmented across curated repositories, computational archives, and datasets extracted from the scientific literature. Inconsistent terminology, measurement units, and identifier schemes limit interoperability and reuse. This study presents an ontology-grounded pipeline that federates two complementary public sources against the eNanoMapper ontology, harmonises records into a canonical schema, preserves evidence in a provenance-aware knowledge graph, and applies gradient-boosted classification with split conformal prediction to a relative antimicrobial hazard-screening task based on MIC/MBC records. The integrated dataset contains 6588 records spanning thirteen material classes and seven harmonised endpoint types, and the knowledge graph contains 2279 nodes and 32,940 typed edges in one weakly connected component. Under five-fold stratified cross-validation, the strongest classifier achieved an area under the receiver-operating-characteristic curve (AUROC) of \(0.819 \pm 0.020\) . On a disjoint test set, split conformal prediction achieved empirical marginal coverage of 0.955, 0.929, and 0.825 against nominal targets of 0.95, 0.90, and 0.80, respectively. A material-grouped cross-validation stress test reduced AUROC to \(0.609 \pm 0.098\) , demonstrating limited generalisation to unseen material classes. Feature attribution and ablation show that categorical experimental context carries most of the discriminative signal, whereas measured physicochemical descriptors contribute comparatively little. The resulting framework therefore combines semantic data integration, explicit provenance, and uncertainty-aware antimicrobial hazard screening while making the limits of cross-material generalisation explicit.

Authors

Institutions

Publication Details

Journal
Scientific Reports
Published
2026-11-18
DOI
https://doi.org/10.1038/s41598-026-73889-7
Primary Topic
Nanoparticles: synthesis and applications
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Ontology-grounded nanotoxicity data federation with a provenance-aware knowledge graph and conformal prediction for antimicrobial hazard screening

Vaisali Chandrasekar, Pankaj Pandey, Ajay Vikram Singh, Purushottam Singh et al.
1 links
Scientific Reports
Nanoparticles: synthesis and applications
article

Ontology-grounded nanotoxicity data federation with a provenance-aware knowledge graph and conformal prediction for antimicrobial hazard screening

Vaisali Chandrasekar, Pankaj Pandey, Ajay Vikram Singh, Purushottam Singh, Mohit Kumar
article en

Abstract

Nanomaterials are increasingly used across industrial, therapeutic, and consumer applications, while their interactions with biological and environmental systems can produce adverse effects. Although nanotoxicological evidence is abundant, it remains fragmented across curated repositories, computational archives, and datasets extracted from the scientific literature. Inconsistent terminology, measurement units, and identifier schemes limit interoperability and reuse. This study presents an ontology-grounded pipeline that federates two complementary public sources against the eNanoMapper ontology, harmonises records into a canonical schema, preserves evidence in a provenance-aware knowledge graph, and applies gradient-boosted classification with split conformal prediction to a relative antimicrobial hazard-screening task based on MIC/MBC records. The integrated dataset contains 6588 records spanning thirteen material classes and seven harmonised endpoint types, and the knowledge graph contains 2279 nodes and 32,940 typed edges in one weakly connected component. Under five-fold stratified cross-validation, the strongest classifier achieved an area under the receiver-operating-characteristic curve (AUROC) of \(0.819 \pm 0.020\) . On a disjoint test set, split conformal prediction achieved empirical marginal coverage of 0.955, 0.929, and 0.825 against nominal targets of 0.95, 0.90, and 0.80, respectively. A material-grouped cross-validation stress test reduced AUROC to \(0.609 \pm 0.098\) , demonstrating limited generalisation to unseen material classes. Feature attribution and ablation show that categorical experimental context carries most of the discriminative signal, whereas measured physicochemical descriptors contribute comparatively little. The resulting framework therefore combines semantic data integration, explicit provenance, and uncertainty-aware antimicrobial hazard screening while making the limits of cross-material generalisation explicit.

Scientific Reports
Amity University (IN), German Federal Institute for Risk Assessment (BfR) (DE), Jharkhand Rai University (IN), Manipal University Jaipur (IN)
Nanoparticles: synthesis and applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon