AcrDetector

AcrDetector identifies anti-CRISPR proteins (Acrs) across entire genomes to enable genome-scale analysis of their distribution and interactions with CRISPR-Cas systems.

Key Features:

• Random Forest Tree-Based Methodology: AcrDetector uses a random forest tree-based classifier leveraging six specific features to predict Acrs.
• Performance Metrics: Reported mean metrics are Accuracy 99.65%, Recall 75.84%, Precision 99.24%, and F1 score 85.97%.
• Robust Validation: Performance was evaluated by multi-round 5-fold cross-validation across 30 different random states.
• Cross-Species Validation: Cross-species validation ranked 71.43% of real Acrs within the top 10 predicted candidates genome-wide.
• Experimental Verification: Applied to recent datasets, AcrDetector recovered three Acrs previously confirmed by experimental methods.
• Recent Transfer Insight: Analysis indicates most Acrs were transferred into host genomes at a relatively recent evolutionary stage.

Scientific Applications:

• Genome-scale Acr discovery: Identify and prioritize candidate Acrs across bacterial and viral genomes for downstream validation.
• CRISPR-Cas regulation studies: Investigate how Acrs modulate the activity of CRISPR-Cas systems.
• Viral-host interaction and evolutionary analysis: Analyze dynamics of Acr transfer and their role in microbial evolution and viral-host interactions.
• Biotechnology and medical research: Inform studies where Acrs impact genome editing and the development of Acr-related tools.

Methodology:

AcrDetector applies a random forest tree-based classifier using six specific features, evaluated by multi-round 5-fold cross-validation across 30 random states, includes cross-species validation ranking real Acrs in top-10 candidates, and was applied to recent datasets to recover three experimentally confirmed Acrs and to analyze timing of Acr transfer into host genomes.