Leveraging heterogeneous data from GHS toxicity annotations, molecular and protein target descriptors and Tox21 assay readouts to predict and rationalise acute toxicity

Table 2 Protein target descriptors exhibiting a significant difference in distribution between binarized toxic set versus nontoxic set

Uniprot	Protein	Class	Cohen’s d	p value	Relevant function [associated pathologies]	Refs.
P00352	Retinal dehydrogenase 1	Oxidoreductase	0.48	2.0 × 10⁻¹⁷	Metabolism processes and oxidation–reduction process [oxidative stress]	[47]
P51449	Nuclear receptor ROR gamma	Nuclear hormone receptor	0.37	1.4 × 10⁻¹³	Cytokine-mediated signaling pathways [apoptosis]	[49]
Q16236	Nuclear factor erythroid 2-related factor 2	Transcription factor	0.36	5.5 × 10⁻¹³	Regulation of cellular redox conditions [oxidative stress]	[48]
Q8IUX4	DNA dC → dU-editing enzyme APOBEC-3F	Hydrolase	0.33	1.5 × 10⁻⁷	DNA mutators participating in the innate immune system [inducing mutations > apoptosis/necrosis]	[50]
P11473 (pidgin)	Vitamin D3 receptor	Nuclear hormone receptor	0.28	1.0 × 10⁻⁸	Ca²⁺ signalling [negative regulation of cell proliferation]	[51]

Enrichment effect size was quantified via Cohen’s d, and the top 5 targets with the largest effect size indicating more association with the toxic set are shown. Only enrichments with a p value below the Bonferroni-corrected threshold equivalent to α = 0.05 were considered. The full table is given in Additional file 11: Table S3. The relevance of these targets to toxicity is explored in the text