Regenie v3.1.2

• Reduction in memory usage for SKAT/SKATO tests
• Bug fix for LOVO with SKAT/ACAT tests
• Improvements for null Firth logistic algorithm to address reported convergence issues
• Bug fix with low INFO variants with approx. Firth which gave BETA=SE=0
• Updated cxxopts library to v3.0

The Linux & Centos7 binaries are statically compiled with GFortran library (so no need to have it installed on the system).

Regenie v3.1.1

• Reduction in memory usage for SKAT/SKATO tests (relevant for large genes like TTN) - added memory due to SKAT tests should be ~2xM^2x8 bytes where M is the largest number of variants in a set (ultra-rare variants collapsed in 1 mask)
• Improvements for logistic regressions algorithms to address reported convergence issues (favor score criterion over log-likelihood fractional change)
• Updated to Eigen 3.4.0 library

Regenie v3.1

• Fixed bug in SKAT/SKATO tests when applying Firth/SPA correction
• Improved SPA implementation by computing both tail probabilities (previous implementation gave more conservative test with high case control imbalance)
• New option --set-singletons to specify which sites to include in singleton masks when building masks by adding a third column in the AAF file (see documentation)
• New option --l1-phenoList to run level 1 model only for a subset of the phenotypes when using --run-l1 (see wiki)
• Fixed bug in joint tests (occurred when some of the burden masks were ignored due to low MAC/failed test and program would crash)
• Fix bug when using apply-rint (introduced in v3.0.3)
• Added additional convergence criterion for ridge logistic regression based on fractional change
• Fix convergence issue for null Firth when fitted probabilities 0/1 occurred
• Fix format mismatch in sum stats file with joint tests (extra NA entry was added when input file had dosages)

Regenie v3.0.3

• Skip BTs where null logistic model fit failed (instead of throwing an error)
• Bug fix for summary stats output file format when using --no-split
• Bug fix for ACAT joint test on burden masks
• Bug fix when nan/inf values are in phenotype/covariate file

Regenie v3.0.1

• Reduce convergence issues with ridge logistic regression in Step 1
• Now able to compile the Regenie binary with Cmake:

mkdir -p build # from main regenie directory 
cd build
cmake ..  # e.g. to add boost iostreams and intel mkl, use `HAS_BOOST_IOSTREAM=1 MKLROOT=/path_to_mkl/ cmake ..`
make

• Note that this release is also available in conda
• Fixed bug in SKAT/ACAT gene-based tests when the variant set is empty (only affected QT run)

Regenie v3.0

New set of features for association testing:

• Extended gene-based tests: SKAT, SKATO, ACATV, ACATO
• Interaction tests to identify GxE or GxG interactions
• Conditional analyses

Additional changes:

• New option --phenoExcludeList to specify list of phenotypes to ignore from the analysis
• New option --covarExcludeList to specify list of covariates to ignore from the analysis
• With --use-null-firth, it is now ok if some traits are absent from the file
• Relaxed null Firth step to only skip traits which failed instead of throwing an error
• Can use .afreq file from PLINK2 as input to --aaf-file
• Improved implementation for Firth logistic regressions (+bug fixes for multi-trait runs)

Check documentation here: https://rgcgithub.github.io/regenie/overview/

Note: Starting from this version, GFortran library should be installed on the system

Regenie v2.2.4

• Bug fix for multi-trait step 1 run with binary traits of different missingness patterns

Regenie v2.2.3

• Bug fix for binary traits for which null logistic regression gives 0/1 fitted probabilities
• Enabled multi-threading for null model fitting with approximate Firth

Regenie v2.2.2

• Bug fix for binary traits for which null logistic regression gives 0/1 fitted probabilities [bug would result in 1e-307 p-values]
• Sex-specific analyses using --sex-specific

This version contains a bug for binary traits. Use version 2.2.3 instead

Regenie v2.2.1

Bug fix for Step 2 with binary traits for which null logistic regression gives 0/1 fitted probabilities [i.e. highly imbalanced or low case counts]

This version contains a bug for binary traits. Use version 2.2.2 instead