Toxcity values of chemical and pharmaceutical compounds are highly important information for environmental and clinical studies to ensure the safe application of newly discovered molecules and industrial chemistry [1,2,3].
We present an advanced graph neural network which can be applied for the HOMO-LUMO gap predictions of molecular entities. This information can be mapped to a toxicity score of the high-throughput screened molecules to retrieve a preliminary analysis of the potential toxicity of the evaluated molecular structures, before performing toxicity studies with animals [4].
The individual steps of the pipeline are compiled as container images in order to develop property prediction pipelines for critical physical properties estimations or the generation of Hamiltonians for quantum chemistry and quantum computing ab-initio pipelines [5].
Mark Nicholas Jones