ToxiVerse

Assay Profiler

This module builds chemical–assay profiles using PubChem bioassay data to fill experimental gaps. You can upload up to 500 chemicals (by PubChem CID). Assays with at least one active compound are selected; mutual information identifies informative assays; then Random Forest models (trained on up to 500 actives and 500 inactives per informative assay) impute missing values.

Upload a file containing PubChem CIDs (one per line) to profile by assay outcomes. Accepted formats: .txt or .csv, single column. If you have other identifiers (e.g., SMILES), use PubChem’s batch converter: ID Exchange

Download Sample CSV

Download Sample TXT

Upload Dataset File (Formats: .txt, .csv; max 500 CIDs) File must be a single column of CIDs. Empty lines are ignored.

What Each Download Option Means

Download Bioprofile: Exports the bioprofile matrix with chemicals as rows and bioassays as columns. Activity values are encoded as 1 (Active/Probe), 0 (Inactive), and −1 (Inconclusive/Unspecified/empty).
Download Heatmap: A clustered heatmap of the bioprofile matrix to spot activity patterns across assays and chemicals. Supports up to 2000 assay columns.
Download Model Metrics: Detailed performance statistics (accuracy, precision, recall, F1-score, ROC AUC) for Random Forest models on top informative assays.
Download Model Metrics Plot: Bar charts summarizing model performance for quick comparison.
Download Filled Data Gaps Matrix: Returns the chemical–assay matrix with missing activity values imputed using trained models.