Possible outputs from running fine-mapping-inf are:

${PREFIX}.V.npz
${PREFIX}.Dsq.npz
${PREFIX}.susieinf.npz
${PREFIX}.susieinf.bgz
${PREFIX}.finemapinf.npz
${PREFIX}.finemapinf.bgz
${PREFIX}.finemapinf.config.bgz

.npz files

Eigen decompositions

.npz is the default file format for saving eigen decompositions of XtX. If --eigen-decomp-prefix is specified, then .V.npz and .Dsq.npz will be saved to files. The arrays can be loaded with

V = numpy.load('${PREFIX}.V.npz')['arr_0']
Dsq = numpy.load('${PREFIX}.Dsq.npz')['arr_0']

Fine-mapping results

User can specify --save-npz to save the fine-mapping results in the form of dictionaries to ${PREFIX}.susieinf.npz and ${PREFIX}.finemapinf.npz.

${PREFIX}.susieinf.npz contains keys:

PIP, mu, omega, ssq, sigmasq, tausq, alpha, cred, lbf, lbf_variable

See susieinf/susieinf.py for definitions of these keys.

${PREFIX}.finemapinf.npz contains keys:

PIP, beta, se, sigmasq, tausq, alpha, models

See finemapinf/finemapinf/__init__.py for definitions of these keys.

.bgz files

The .bgz files are bgzipped tab delimited text files.

.susieinf.bgz file contains columns:

• rsid, chromosome, position, allele1, allele2, maf: basic variant info
• beta, se, p: marginal summary statistics
• prob: posterior inclusion probability for each variant
• alpha{1..L}, mu{1..L}, omega{1..L}, lbf_variable{1..L}: component-wise parameters. Same definition as in susieinf/susieinf.py
• alpha, tausq, sigmasq: see susieinf/susieinf.py for definitions
• post_mean: computed as sum(alpha{1..L}*mu{1..L},axis=1)+alpha, posterior mean effect size (including infinitesimal effects)
• cs: credible set ID, -1 means variant is not in any credible set

.finemapinf.bgz file contains columns:

• rsid, chromosome, position, allele1, allele2, maf: basic variant info
• beta, se, p: marginal summary statistics
• prob: posterior inclusion probability for each variant
• post_mean_cond, post_sd_cond: posterior conditional mean and standard deviation
• alpha, tausq, sigmasq: see finemapinf/finemapinf/__init__.py for definitions
• post_mean: posterior mean, computed as post_mean_cond*prob

.finemapinf.config.bgz file contains columns:

• config: causal configuration; 0-indexed; empty list represent the null model
• prob: posterior probability of configuration being causal
• prob: posterior probability