摘要：Transcription factors (TFs) are proteins controlling the fow of genetic information by regulating cel‑ lular gene expression. A better understanding of TFs in a bacterial community context may open novel revenues for exploring gene regulation in ecosystems where bacteria play a key role. Here we describe PredicTF, a platform sup‑ porting the prediction and classifcation of novel bacterial TF in single species and complex microbial communities. PredicTF is based on a deep learning algorithm. Results: To train PredicTF, we created a TF database (BacTFDB) by manually curating a total of 11,961 TF distributed in 99 TF families. Five model organisms were used to test the performance and the accuracy of PredicTF. PredicTF was able to identify 24–62% of the known TFs with an average precision of 88% in our fve model organisms. We demon‑ strated PredicTF using pure cultures and a complex microbial community. In these demonstrations, we used (meta) genomes for TF prediction and (meta)transcriptomes for determining the expression of putative TFs. Conclusion: PredicTF demonstrated high accuracy in predicting transcription factors in model organisms. We pre‑ pared the pipeline to be easily implemented in studies profling TFs using (meta)genomes and (meta)transcriptomes. PredicTF is an open-source software available at https://github.com/mdsufz/PredicTF.
