Bioinformatics and chemogenomics have seen recent advancements that offer promising avenues to expedite the discovery of small-molecule regulators that influence cell development (Fawcett, Pattern Recognit Lett 27:861–874, 2006). By integrating extensive genomics and molecular data sources with robust deep learning techniques, we can potentially transform predictive biology. Our preliminary study results highlight the effectiveness of deep learning models in leveraging molecular and genomic descriptors to screen for novel drug candidates that can positively impact gene expression. These models hold great promise in advancing the development of innovative cancer therapies and driving precision oncology initiatives. In this investigation, we harnessed molecular fingerprint descriptors and gene descriptors, derived from Gene Ontology (GO) terms, to construct and train deep neural networks (DNNs). Our purpose was to predict diverse gene regulation outcomes. We used datasets sourced from the LINCS database to train our models. To create our training datasets, we segmented the differential gene expression values from the LINCS database based on a threshold of 5%. Values exceeding the 95% threshold were classified as active gene up-regulation, while those below the 5% threshold were categorized as active gene down-regulation. During our research, we utilized deep learning models that included an input layer, hidden layers, and an output layer.