The indoor chemical exposome across five Asian regions: socio-economic associations, network structure, and toxicity prioritization

Abstract

This dataset accompanies the manuscript "The indoor chemical exposome across five Asian regions: socio-economic associations, network structure, and toxicity prioritization" (Environment International). It contains all data and analysis code required to reproduce the results, figures, and supplementary tables of the study: - data/ : 12 cleaned files, comprising class-specific concentration matrices (ng/g) for 149 indoor dust samples across eight chemical classes, a 149 x 54 complete-case matrix used for network and heatmap analyses, a 149 x 82 feature matrix used for machine learning, sample-to-region metadata, and the compound dictionary (CAS number, DTXSID). - code/ : 21 analysis scripts (R and Python) covering data cleaning and LOQ/2 imputation, descriptive and multivariate analyses (heatmap, t-SNE/UMAP), Mantel tests, XGBoost/RF/SVR with SHAP interpretation, co-occurrence network construction with FDR correction, health risk assessment (ADD/HQ/HI) and toxicity prioritization, and all sensitivity analyses. - README.txt: full documentation, including column definitions, method notes, software versions, and a step-by-step reproduction order mapping every figure and supplementary table to its generating script. Samples were collected from educational settings in Sri Lanka, Nepal, Malaysia, Indonesia, and Guangxi (China). The design is convenience-based and unbalanced; the derived toxicity priorities are screening-level results for this sample set and should not be generalised to national or income-level priorities.