Release v1.0.0

MR-link-2 Release v1.0.0

This is the main release for MR-link-2.

At this point we are confident in the results of the method. Particularly based on it's application to metabolites, complex traits and gene expression, as well as based on feedback from collaborators, we believe the method is ready for broader consumption.

We also invested a lot in testing that the method remains sound even with a wide variety of data formats. And inputs.

Based on collaborator feedback, we provide an R version that provides the same results as the python version, but does not have the same preprocessing as the python version. So when using this, beware of preprocessing pitfalls, which the python version has less of.

Changes:

• Code has been peer reviewed and results are stable and reasonable.
• License has been changed to MIT for more broader use
• We provide an R version of the method, which is tested to provide the same results as the python version
• Integration testing and unit testing

MR-link-2 release v0.1.2

Release of MR-link-2 v0.1.2

This release contains a bugfix and a large refactor that increases the speed of processing.

This release adds a little bit of code-complexity, but allows for fast processing of multiple outcomes the same exposure. We have been able to test 10 million two-phenotype comparisons in acceptable time (days rather than months) on a relatively small cluster.

Features

• MR-link-2 now accepts the GWAS summary statistics file format as well as the originally defined format.
• MR-link-2 now accepts files in the parquet format for fast and random reading.
• We now accept multiple outcome summary statistics files, which will reduce the times that the LD matrix needs to be read, which can be a large time-sink.

Bug fixes

Fixed a (stupid) bug where the SNP ordering wasn't passed to a function correctly when there are missing genotypes.

v0.1.1

This is the first release for MR-link-2.

Speed improvements:

• Wrote a plink style bed parser to read in a correlation matrix faster. This will improve running times of phenotypes with a lot of genetically associated regions.

Bugfixes:

• Now, MR-link-2 will check to ensure that the alleles for the LD matrix and the summary statistics are the same. This will not greatly change your analysis, but will generally lead to more stable sigma_y estimates.

Full Changelog: v0.1.0...v0.1.1