bio-genome-engineering-off-target-prediction

Cas-OFFinder (CLI)

Cas-OFFinder searches genomes for potential off-target sites allowing mismatches.

## Position-specific mismatch penalties

CFD_MISMATCH_SCORES = {
    # (position, ref_nt, target_nt): penalty_multiplier
    # Position 1 = PAM-proximal, Position 20 = PAM-distal
    # Values < 1 indicate reduced cutting
    (1, 'C', 'A'): 0.5, (1, 'C', 'G'): 0.7, (1, 'C', 'T'): 0.3,
    (1, 'G', 'A'): 0.4, (1, 'G', 'C'): 0.6, (1, 'G', 'T'): 0.3,
    # ... (full matrix has all 20 positions x 12 mismatch types)
    (20, 'C', 'A'): 0.9, (20, 'C', 'G'): 0.95, (20, 'C', 'T'): 0.85,
}

## PAM mismatch penalties

CFD_PAM_SCORES = {
    'AGG': 0.26, 'CGG': 0.11, 'TGG': 0.02,  # First position
    'GAG': 0.07, 'GCG': 0.03, 'GTG': 0.02,  # Second position
    'GGA': 0.01, 'GGC': 0.01, 'GGT': 0.01,  # Third position
}

def calculate_cfd_score(guide, off_target, pam='NGG'):
    '''Calculate CFD score for an off-target site

    CFD score interpretation:
    - 1.0: Perfect match (on-target)
    - >0.5: High probability of cleavage (concerning)
    - 0.1-0.5: Moderate probability
    - <0.1: Low probability (likely acceptable)
    '''
    score = 1.0

    # Apply mismatch penalties
    for i, (g, t) in enumerate(zip(guide, off_target[:20]), 1):
        if g != t:
            key = (i, g, t)
            penalty = CFD_MISMATCH_SCORES.get(key, 0.5)  # Default 0.5
            score *= penalty

    # Apply PAM penalty if not NGG
    off_pam = off_target[20:23]
    if off_pam != 'NGG' and off_pam in CFD_PAM_SCORES:
        score *= CFD_PAM_SCORES[off_pam]

    return score