Deep learning methods have been developed to predict gene expression levels across various cell and tissue types based on DNA-binding sites of protein identified by ChIP-seq. However, the experimental identification of DNA/RNA-binding sites and the functional characterization of certain proteins remain challenging. Here I predicted the regulatory target genes and functional roles of DNA/RNA-binding proteins using gene co-expression data. Gene co-expression correlations are derived from gene expression profiles across cells and tissues and represent interactions between a DNA/RNA-binding protein and other genes. I used these interactions to predict gene expression levels by partially replacing the input data in the deep learning analysis. Functional annotations statistically overrepresented in the predicted target genes of ten DNA/RNA-binding proteins examined in this study were consistent with their known biological roles. Furthermore, several annotations not present in current gene function databases were supported by findings in previously published studies.