BacDive AI models

This is a small command line tool bundled with selected BacDive-AI models to predict bacterial phenotypes from InterPro (Pfam) annotations. Please see the following Preprint for more information: https://doi.org/10.1101/2024.08.12.607695

Installation instructions

This repository comes with zipped models because of their large file size. Please unzip the models into the /models path before using them.

Furthermore, the following requirements must be met:

• Python v3.9 or higher
• scikit-learn v1.4.0

This package comes with an example InterPro annotation of the strain Actinomyces dentalis DSM 19115. The annotated file in TSV format is named as 1120941.3.faa.tsv. The genome is derived from BV-BRC (formerly PATRIC).

For reference, the used command to generate the Pfams with InterProScan was:

interproscan.sh -i 1120941.3.faa -f tsv -d ./ -appl Pfam

To start the prediction of a single trait, you can use the following example code in the project folder:

python predict.py gram-positive 1120941.3.faa.tsv

For the aforementioned call, you can select one of the following values:

Value	Trait
acidophile	Acidophilic
gram-positive	Gram-positive
spore-forming	Spore-forming
aerobic	Aerobic
anaerobic	Anaerobic
thermophile	Thermophilic
psychrophile	Psychrophilic
motile2+	Flagellated motility

You can also choose all to see the complete list of predicted values:

python predict.py all 1120941.3.faa.tsv

You will get the prediction (true or false) for each of the trait together with a confidence score. The output should look like this:

Acidophilic: False (98.5%)
Gram-positive: True (98.93%)
Spore-forming: False (96.47%)
Aerobic: False (93.91%)
Anaerobic: True (63.54%)
Thermophilic: False (99.77%)
Psychrophilic: False (99.96%)
Flagellated motility: False (99.85%)