Skip to content

cojo

Wrapper for COJO.

conditional_selection(locus, p_cutoff=5e-08, collinear_cutoff=0.9, window_size=10000000, maf_cutoff=0.01, diff_freq_cutoff=0.2)

Perform conditional selection on the locus using COJO method.

Parameters

locus : Locus The locus to perform conditional selection on. Must contain summary statistics and LD matrix data. p_cutoff : float, optional The p-value cutoff for the conditional selection, by default 5e-8. If no SNPs pass this threshold, it will be relaxed to 1e-5. collinear_cutoff : float, optional The collinearity cutoff for the conditional selection, by default 0.9. SNPs with LD correlation above this threshold are considered collinear. window_size : int, optional The window size in base pairs for the conditional selection, by default 10000000. SNPs within this window are considered for conditional analysis. maf_cutoff : float, optional The minor allele frequency cutoff for the conditional selection, by default 0.01. SNPs with MAF below this threshold are excluded. diff_freq_cutoff : float, optional The difference in frequency cutoff between summary statistics and reference panel, by default 0.2. SNPs with frequency differences above this threshold are excluded.

Returns

pd.DataFrame The conditional selection results containing independently associated variants with columns including SNP identifiers, effect sizes, and conditional p-values.

Warnings

If no SNPs pass the initial p-value cutoff, the threshold is automatically relaxed to 1e-5 and a warning is logged.

If AF2 (reference allele frequency) is not available in the LD matrix, a warning is logged and frequency checking is disabled.

Notes

COJO (Conditional and Joint analysis) performs stepwise conditional analysis to identify independently associated variants at a locus. The method:

  1. Identifies the most significant SNP
  2. Performs conditional analysis on remaining SNPs
  3. Iteratively adds independently associated SNPs
  4. Continues until no more SNPs meet significance criteria

The algorithm accounts for linkage disequilibrium patterns and helps distinguish truly independent signals from those in LD with lead variants.

Reference: Yang, J. et al. Conditional and joint multiple-SNP analysis of GWAS summary statistics identifies additional variants influencing complex traits. Nat Genet 44, 369-375 (2012).

Examples

Basic conditional selection

results = conditional_selection(locus) print(f"Found {len(results)} independent signals") Found 3 independent signals

With custom thresholds

results = conditional_selection( ... locus, ... p_cutoff=1e-6, ... maf_cutoff=0.05 ... ) print(results[['SNP', 'b', 'se', 'p']]) SNP b se p 0 rs123456 0.15 0.025 1.2e-08 1 rs789012 -0.08 0.020 4.5e-07

Source code in credtools/cojo.py 
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
def conditional_selection(
    locus: Locus,
    p_cutoff: float = 5e-8,
    collinear_cutoff: float = 0.9,
    window_size: int = 10000000,
    maf_cutoff: float = 0.01,
    diff_freq_cutoff: float = 0.2,
) -> pd.DataFrame:
    """
    Perform conditional selection on the locus using COJO method.

    Parameters
    ----------
    locus : Locus
        The locus to perform conditional selection on. Must contain summary statistics
        and LD matrix data.
    p_cutoff : float, optional
        The p-value cutoff for the conditional selection, by default 5e-8.
        If no SNPs pass this threshold, it will be relaxed to 1e-5.
    collinear_cutoff : float, optional
        The collinearity cutoff for the conditional selection, by default 0.9.
        SNPs with LD correlation above this threshold are considered collinear.
    window_size : int, optional
        The window size in base pairs for the conditional selection, by default 10000000.
        SNPs within this window are considered for conditional analysis.
    maf_cutoff : float, optional
        The minor allele frequency cutoff for the conditional selection, by default 0.01.
        SNPs with MAF below this threshold are excluded.
    diff_freq_cutoff : float, optional
        The difference in frequency cutoff between summary statistics and reference panel,
        by default 0.2. SNPs with frequency differences above this threshold are excluded.

    Returns
    -------
    pd.DataFrame
        The conditional selection results containing independently associated variants
        with columns including SNP identifiers, effect sizes, and conditional p-values.

    Warnings
    --------
    If no SNPs pass the initial p-value cutoff, the threshold is automatically
    relaxed to 1e-5 and a warning is logged.

    If AF2 (reference allele frequency) is not available in the LD matrix,
    a warning is logged and frequency checking is disabled.

    Notes
    -----
    COJO (Conditional and Joint analysis) performs stepwise conditional analysis
    to identify independently associated variants at a locus. The method:

    1. Identifies the most significant SNP
    2. Performs conditional analysis on remaining SNPs
    3. Iteratively adds independently associated SNPs
    4. Continues until no more SNPs meet significance criteria

    The algorithm accounts for linkage disequilibrium patterns and helps
    distinguish truly independent signals from those in LD with lead variants.

    Reference: Yang, J. et al. Conditional and joint multiple-SNP analysis of GWAS
    summary statistics identifies additional variants influencing complex traits.
    Nat Genet 44, 369-375 (2012).

    Examples
    --------
    >>> # Basic conditional selection
    >>> results = conditional_selection(locus)
    >>> print(f"Found {len(results)} independent signals")
    Found 3 independent signals

    >>> # With custom thresholds
    >>> results = conditional_selection(
    ...     locus,
    ...     p_cutoff=1e-6,
    ...     maf_cutoff=0.05
    ... )
    >>> print(results[['SNP', 'b', 'se', 'p']])
        SNP           b        se         p
    0   rs123456   0.15   0.025   1.2e-08
    1   rs789012  -0.08   0.020   4.5e-07
    """
    sumstats = locus.sumstats.copy()
    sumstats = sumstats[
        [
            ColName.SNPID,
            ColName.EA,
            ColName.NEA,
            ColName.BETA,
            ColName.SE,
            ColName.P,
            ColName.EAF,
        ]
    ]
    sumstats.columns = ["SNP", "A1", "A2", "b", "se", "p", "freq"]
    sumstats["N"] = locus.sample_size
    if p_cutoff < 1e-5 and len(sumstats[sumstats["p"] < p_cutoff]) == 0:
        logger.warning("No SNPs passed the p-value cutoff, using p_cutoff=1e-5")
        p_cutoff = 1e-5

    ld_matrix = locus.ld.r.copy()
    ld_freq: Optional[pd.DataFrame] = locus.ld.map.copy()
    if ld_freq is not None and "AF2" not in ld_freq.columns:
        logger.warning("AF2 is not in the LD matrix.")
        ld_freq = None
    elif ld_freq is not None:
        ld_freq = ld_freq[["SNPID", "AF2"]]
        ld_freq.columns = ["SNP", "freq"]
        ld_freq["freq"] = 1 - ld_freq["freq"]
    c = COJO(
        p_cutoff=p_cutoff,
        collinear_cutoff=collinear_cutoff,
        window_size=window_size,
        maf_cutoff=maf_cutoff,
        diff_freq_cutoff=diff_freq_cutoff,
    )
    c.load_sumstats(sumstats=sumstats, ld_matrix=ld_matrix, ld_freq=ld_freq)  # type: ignore
    cojo_result = c.conditional_selection()
    return cojo_result