Introduction to AOminer

AOminer is the first ever automated tool for reliable screening of anti-CRISPR operons (AOs) using machine learning HMM algorithms. Using the built-in two-state HMM trained based on known AOs, AOminer is able to screen for potential AOs from both genomic data and individual operons. AOminer holds an accuracy of 89% from our performance tests, and outperformed all other machine learning based anti-CRISPR prediction tools in terms of AO prediction. In addition, AOminer also searches the input genome/operon for Acr homologs using published Acr proteins, potential Aca proteins containing HTH domains in predicted AOs, CRISPR-Cas systems using (CRISPRCasTyper) and an in-house Self-targeting spacer (STS) search tool, prophage regions using (VIBRANT). All information will be integrated with the predicted AOs, providing users with detailed information vital to prediction assessment.

AOminer Workflow

Step 1) The input FAA will be scanned for short-gen operons (SGOs) with the following criteria: (i) All genes < 200aa (if Acr homolog present in SOG and homolog is >200aa, then all genes < Acr homolog length); (ii) All intergenic distances < 250bp; (iii) All genes on the same strand. If input is a single operon, then this step is bypassed.
Step 2) The extracted SGOs or input operon will be annotated with the built-in dbPFhmm, of which contain HMMs of protein families found in abundance within known AOs.
Step 3) The annotated operons will go through the built-in two-state HMM. Each operon will be assigned a prediction score, the ones that passes the set thresholds will be considered as a predicted AO. The predicted AOs will then be annotated with the Pfam database using pfamscan.
In addition to the identification of anti-CRISPR operons, AOminer will also scan the input fna file for prophages, CRISPR-Cas systems & self-targeting spacers (STSs), Acr homologs, and potential Acas. If input is a single operon, only Acr homologs and potetnial Aca proteins will be searched.