Output files¶

The output files consist of eight types:

*.log
*.cs.tsv
*.alphas.tsv
*.weights.tsv
*.corr.tsv
*.her.tsv
*.cv.tsv
*.all.results.npy

Users can output them as compressed files by specifying --compress

Logger¶

SuShiE by default has a *.log file that keep tracks of inference process.

Credible set¶

SuShiE by default outputs a *.cs.tsv file that tracks the SNPs in the credible sets.

If --meta and --mega are specified (see meta and mega SuShiE), it will output *.meta.cs.tsv and *.mega.cs.tsv, respectively, to track the SNPs in the credible sets inferred by meta SuShiE and mega SuShiE.

For *.meta.cs.tsv, it will row-bind the output for single-ancestry SuShiE differed by column ancestry.

Column	Type	Examples	Notes
SNPIndex	Integer	1, 33, 77	The SNP unique index identifier. It matches the order of entries in the original genotype data counting start from 0.
chrom	Integer	1, 23	The chromosome number.
snp	String	rs12345	The SNP unique identifier (e.g., rs ID). It matches the entries in the original genotype data.
pos	Integer	123456	The SNP position on the chromosome. It matches the entries in the original genotype data.
a0	String	A	The non-counting allele. It matches the entries in the original genotype data.
a1	String	G	The counting allele. If the counting allele is G, and genotype is GG, then it is coded as 2. It matches the entries in the original genotype data.
CSIndex	Integer	1, 2	The credible set unique index.
alpha	Float	0.8	The posterior probability of SNPs to be causal in the corresponding credible set (\(\alpha_{l,j}\) in the model).
c_alpha	Float	0.95	The cumulative posterior probability of SNPs to be causal in the descending order. This decides which SNPs are included in the credible sets.
pip_all	Float	0.95	The posterior inclusion probability (\(\text{PIP}_j\) in the model) calculated across \(L\) credible sets. For `*.meta.cs.tsv`, it will have an additional column called `meta_pip_all`.
pip_cs	Float	0.95	The posterior inclusion probability (\(\text{PIP}_j\) in the model) calculated across credible sets that are kept after pruning based on purity. For `*.meta.cs.tsv`, it will have an additional column called `meta_pip_cs`.
trait	String	GeneABC	The trait, tissue, or gene name.
n_snps	Integer	500	The number of total SNPs in the inference.
ancestry	String	sushie, mega, ancestry_1	The inference method for this credible set.

Full credible set with alphas¶

By specifying --alphas, SuShiE outputs a *.alphas.tsv file that tracks all the SNPs' PIP, \(\alpha\) (see the model), and purity across all \(L\).

If --meta and --mega are specified (see meta and mega SuShiE), it will output *.meta.alphas.tsv and *.mega.alphas.tsv, respectively, to track the information inferred by meta SuShiE and mega SuShiE.

For *.meta.alphas.tsv, it will row-bind the output for single-ancestry SuShiE differed by column ancestry.

Column	Type	Examples	Notes
SNPIndex	Integer	1, 33, 77	The zero-based SNP index in the original genotype data.
chrom	Integer	1, 23	The chromosome number.
snp	String	rs12345	The SNP identifier from the original genotype data.
pos	Integer	123456	The SNP position on the chromosome.
a0	String	A	The non-counting allele.
a1	String	G	The counted allele; an `a1/a1` genotype is coded as 2.
alpha_l1	Float	0.8	The posterior causal probability for the first credible set. Additional effects add corresponding columns.
in_cs_l1	Integer	0, 1	Whether the SNP belongs to the first credible set.
purity_l1	Float	0.634	The sample-size-weighted mean purity across ancestries.
kept_l1	Integer	0, 1	Whether the credible set remains after purity filtering.
log_bf_l1	Float	100.23	The log Bayes factor for the first credible set.
trait	String	GeneABC	The trait, tissue, or gene name.
n_snps	Integer	500	The number of SNPs in the inference.
purity_threshold	Float	0.5	The purity threshold used to prune credible sets.
ancestry	String	sushie, mega, ancestry_1	The inference method for this credible set.

Prediction weights¶

SuShiE by default outputs a *.weights.tsv file that contains the prediction weights, PIPs, and whether in CS, across all the fine-mapped SNPs.

If --meta and --mega are specified (see meta and mega SuShiE), it will output *.meta.weights.tsv and *.mega.weights.tsv, respectively.

Column	Type	Examples	Notes
SNPIndex	Integer	1, 33, 77	The SNP unique index identifier. It matches the order of entries in the original genotype data counting start from 0.
chrom	Integer	1, 23	The chromosome number.
snp	String	rs12345	The SNP unique identifier (e.g., rs ID). It matches the entries in the original genotype data.
pos	Integer	123456	The SNP position on the chromosome. It matches the entries in the original genotype data.
a0	String	A	The non-counting allele. It matches the entries in the original genotype data.
a1	String	G	The counting allele. If the counting allele is G, and genotype is GG, then it is coded as 2. It matches the entries in the original genotype data.
trait	String	GeneABC	The trait, tissue, or gene name.
ancestry1_sushie_weight	Float	1.3	The ancestry-specific SNP prediction weights inferred by SuShiE. For `*.meta.weights.tsv`, it will have `ancestry1_single_weight` (It will have extra columns depending on the number of ancestries). If `--mega`, it will have `mega_weight` for all ancestries.
sushie_pip_all	Float	0.95	The posterior inclusion probability (\(\text{PIP}_j\) in model) for all the SNPs calculated across \(L\) credible sets. (`.cs.tsv` only contains the PIPs of SNPs that are only in the credible sets). For `.meta.weights.tsv`, it will have `ancestry1_single_pip`, `meta_pip_all` (It will have extra columns depending on the number of ancestries). For `*.mega.weights.tsv`, it will have `mega_pip_all`.
sushie_pip_cs	Float	0.95	The posterior inclusion probability (\(\text{PIP}_j\) in model) for all the SNPs calculated across credible sets that are kept after purning based on purity. (`.cs.tsv` only contains the PIPs of SNPs that are only in the credible sets). For `.meta.weights.tsv`, it will have `ancestry1_single_pip`, `meta_pip_cs` (It will have extra columns depending on the number of ancestries). For `*.mega.weights.tsv`, it will have `mega_pip_cs`.
sushie_cs_index	Integer	0, 1, ..., \(L\)	The credible set index where the SNPs fall into. 0 means no credible sets contain this SNP. For `.meta.weights.tsv`, it will have `ancestry1_cs_index`(It will have extra columns depending on the number of ancestries). For `.mega.weights.tsv`, it will have `mega_cs_index`.
n_snps	Integer	500	The number of total SNPs in the inference.

Effect-size correlation¶

SuShiE by default outputs a *.corr.tsv file that contains the estimated effect size covariance matrix for each output credible set (after pruning for purity). For results of all \(L\) credible sets, see NumPy results.

Column	Type	Examples	Notes
trait	String	GeneABC	The trait, tissue, or gene name.
CSIndex	Integer	1, 2	The credible set unique index. It depends on `--L` and puring after purity.
ancestry1_est_var	Float	1.34	The inferred effect size variance (the posterior estimate for \(\sigma^2_{i,b}\) in the model) for ancestry 1. It depends on the number of ancestry. One estimate for each credible set.
ancestry1_ancestry2_est_covar	Float	2.56	The inferred effect size covariance between ancestry 1 and ancestry 2. It depends on the number of pairs of ancestries. One estimate for each credible set.
ancestry1_ancestry2_est_corr	Float	0.8	The inferred effect size correlation (the posterior estimate for \(\rho\) in the model) between ancestry 1 and ancestry 2. It depends on the number of pairs of ancestries. One estimate for each credible set.

Heritability¶

By specifying --her, SuShiE outputs a *.her.tsv file that tracks the heritability analysis results for each ancestry.

It contains two rounds of heritability estimation:

Using all the SNPs.
Using the SNPs in the credible set (only if SuShiE outputs non-empty credible sets after pruning for purity).

Column	Type	Examples	Notes
ancestry	Integer	1, 2	The ancestry index.
genetic_var	Float	1.32	The variance of genetic components contributing to the complex traits. `s_genetic_var`, which is estimated only from the SNPs in the credible sets, will be appended if credible sets are not empty after pruning for purity.
h2g	Float	0.23	The narrow-sense cis-heritability of the traits based on limix definition. This includes the variance of the fixed effects.
lrt_stats	Float	-123.23	The likelihood ratio test statistics compared the linear mixed effects model to the fixed effects model (no genetic variance). `s_lrt_stats`, which is estimated only from the SNPs in the credible sets, will be appended if credible sets are not empty after pruning for purity.
p_value	Float	-123.23	The \(p\) value for the likelihood ratio test statistics based on chi-square distribution with 1 dof. `s_p_value`, which is estimated only from the SNPs in the credible sets, will be appended if credible sets are not empty after pruning for purity.
trait	String	GeneABC	The trait, tissue, or gene name.

Cross-validation¶

By specifying --cv, SuShiE outputs a *.cv.tsv file that contains the results from cross validation (see the user manual for how we compute the \(r^2\)).

Column	Type	Examples	Notes
ancestry	Integer	1, 2	The ancestry index.
rsq	Float	0.9	\(r^2\) between predicted and measured expressions from cross-validations.
p_value	Float	0.23	The \(p\) value for the \(r^2\).
N	Integer	200	The sample size for SuShiE inference.
trait	String	GeneABC	The trait, tissue, or gene name.

NumPy results¶

By specifying --numpy, SuShiE outputs a *.all.results.npy file that contains all the results from inference and snp information. It can only be read by python numpy package.