Significance Identification of transcription factors (TFs) is a starting point for the analysis of transcriptional regulatory systems of organisms. Here, we report the development of DeepTFactor, a deep learning-based tool that predicts TFs using protein sequences as inputs. We interpreted the reasoning process of DeepTFactor, confirming that DeepTFactor inherently learned DNA-binding domains of TFs. DeepTFactor predicted 332 TFs of E. coli K-12 MG1655, and three of them were experimentally validated by identifying genome-wide binding sites with ChIP-exo experiments. We provide DeepTFactor as a stand-alone program for researchers to analyze their own protein sequences of interest. It will serve as a useful tool for understanding the regulatory systems of organisms.