PredicTF: prediction of bacterial transcription factors in complex microbial communities using deep learning

Table 1 PredicTF performance, accuracy for experimentally validated Transcription Factors (Accuracy EV), and accuracy for putative Transcription Factors (Accuracy PU) in genomes of model organisms. We removed the sequences from the genus and/or family of the different model organisms from our TF database (BacTFDB) before model training to reduce the chances of false positives (i.e., the presence of identical sequences in the training dataset)

Organism	Performance^a (%)	Accuracy^b EV^b (%)	Model^c	Accuracy PU^d
E. coli k12	33.33	94.44	PredicTF-no- Escherichia	85.71
E. coli k12	31.27	95.12	PredicTF-no- Enterobacteriaceae	85.71
B. subtillis	24.26	87.76	PredicTF-no- Bacillus	100
C. crescentus	34.36	96.00	PredicTF-no- Caulobacter	-^e
P. fluorescens	46.44	98.69	PredicTF-no- Pseudomonas	-
A.vinelandii	62.28	98.43	PredicTF-no- Azotobacter	-

^aPerformance was calculated by the ratio of the total number of TFs predicted by PredicTF (Predicted TFs) to the total number of proteins annotated as TFs in NCBI (Annotated TFs) multiplied by 100
^bAccuracy EV was determined by the ratio of the total number of TFs predicted by PredicTF in agreement with NCBI annotation (TFs predicted correctly) to the total number of TFs predicted by PredicTF (TFs predicted) multiplied by 100
^cPrecicTF Model used for the prediction of TF to the specific organism
^dAccuracy TU was determined by the total number of putative TFs predicted correctly divided by putative TFs predicted multiplied by 100; Putative TFs predicted correctly is the total number of putative TFs predicted correctly by PredicTF in agreement with NCBI annotation; and, Putative TFs predicted is the total number of putative TFs predicted by PredicTF
^eCurrently there are no putative annotated TFs described in the genome of C. crescentus, P. fluorescens and A.vinelandii

ISSN: 2524-6372