Remeta v0.9.0

• Improvements to saddlepoint approximation (SPA) in remeta gene to control type 1 error for unbalanced binary traits. This is a breaking change from 0.7.1. This change primarily impacts SKATO and ACATV.
  • The old options --spa-pval and --spa-ccr, which applied a mask-based SPA to burden tests and SKATO, have been replaced by per test options --<burden,skato>-mask-spa-<pval,ccr>. The default parameter to apply the mask-based SPA has been lowered from a case-control ratio of 1:100 to 1:50. This will change p-values of burden tests and SKATO.
  • Added new options to apply a variant-based SPA and --<burden,skato,acatv>-sv-spa-<pval,ccr>. This will result in changes to p-values of SKATO below a case-control ratio of 1:50 and ACATV below a case-control ratio of 1:10.
• Improved estimation of burden effect sizes by recomputing effect sizes from the score statistics and standard errors. This change removes redundant computation, ensures the per cohort burden tests are consistent with the final meta-analysis result, and ensures that the burden test statistic is consistent with the estimated effect sizes.
• Added support for tabixed annotation files in addition to the custom index built by remeta index-anno.
• Updated default --burden-singleton-def parameter from within to across.
• Fixed a bug in remeta merge where fractional genotype counts were truncated.
• Fixed a bug where LOG10P=NA in the info field would cause pvma to crash.
• Allow single variants with different Model columns to be meta-analyzed together.
• Remove lower bound on the Pval column. P-values that underflow will now be reported exactly.
• Added flag --two-sided to PVMA to compute two-sided p-values when using Stouffer's method and the Effect column is not NA.
• Reduced runtime of remeta gene by removing expensive AAF calculation. The AAF column now reports the AAF computed from genotype counts for gene-based tests.
• Added option --write-variant-aaf to output AAFs used to construct gene-based tests.