There are over 350,000 registered chemicals and chemical combinations in use today globally. This is a public health concern and an active area of research, as for majority of these chemicals no scientific data is available on potential adverse effects on human health. Both U.S. Environmental Protection Agency (EPA) and Word Health Organization (WHO) have listed heavy metals, lead (Pb), mercury (Hg), cadmium (Cd) and arsenic (As) among the top chemicals of public health concern. Adverse health effects, induced by heavy metals and other non-heavy metal chemicals include neurodegenerative diseases (Alzheimer’s disease, Parkinson’s disease), cognitive decline, behavioral problems, kidney diseases, cancer and cardiovascular diseases. Thus, it is important to detect not only active chemical exposure but also past chemical exposures. In this paper differential gene expression (DEG) analysis and machine learning (ML) were combined to identify differentially expressed genes (DEGs) or heavy metal toxicity signature genes that were used as features in ML to classify test samples into heavy metal and non-heavy metal control groups. From NIH-GEO, RNA-seq gene expression data from a total of 827 human neuronal cell culture samples treated with 87 different chemicals were downloaded and normalized. Two groups of DEGs consisting of 80 genes (consensus of limma, edgeR and simple DEG analysis) and 879 genes (consensus of atleast 2 of the three DEG methods limma, edgeR and simple) were identified and designated as heavy metal biomarkers. The heavy metal biomarker gene sets were enriched with metal metabolism gene ontology, kidney disease and cancer diseases genes. Comparison of different ML models built with 80 DEGs and 879 DEGs showed that Logistic Regression and Support Vector Machine (SVM) were accurate (>90% success in classifying test samples into heavy metal and non-heavy metal groups) for both 80 DEG and 879 DEG features. In this paper, a combined DEG analysis and ML pipeline has been developed that can successfully detect heavy metal exposure from gene expression data. This pipeline can be applied for identification of chemical exposure, which is the first step for developing a treatment plan for patients exposed to toxic chemicals. Code Availability: Computational code with html or pdf rendering showing input and output of code chunks is available as a git local repository at https://icedrive.net/s/h3P65RbNvf5Dh8yT1DabXxyNgWg6 with all files and as a git remote repository at https://gitlab.com/smukher2/pbothers_rnaseq_ml_feb2023 with large files ignored or removed.