Linear (mixed) model effects to odds ratios

Whole file transformation

Choose file delimeter

Individual locus test


This Shiny app accompanies the manuscript titled 'Transformation of summary statistics from linear mixed model association on all-or-none traits to odds ratio'. This application is designed to map regression coefficients from a linear (mixed) model (LMM) to the odds ratio from genome-wide association studies (GWAS) on disease traits. It has been shown to be effective at mapping effects generated from a linear mixed model GWAS to the odds ratio. This allows for a comparison between effects generated from logistic regression from other GWAS studies.

Reference allele frequencies can be used if the user's summary statistics do not have sample allele frequencies. If this is the case then please add a column named 'FREQ' with the reference allele frequency for the effect allele.

Data formats

The application supports most summary statistics formats from LMM GWAS and requires at least the allele frequency for each SNP, the prevalence in the sample k (written as K in the header of the file), and the estimated effect. The below snapshot shows an example format from GCTA. Note that not all columns are required. PLEASE LIMIT WHOLE FILE TRANSFORMATIONS TO 1M VARIANTS.

If transformations for more than 1M variants are required please download the R function below, which can be used as a standalone function within R.


NB: This function has the option to invoke the standard error (SE) version of the set of transformations derived in the paper above. This transformation was derived for users that may not have an allele frequency, or a good allele frequency reference, for their variants. If you are interested in using the SE transformation we suggest a more thorough reading of the manuscript above and expect more NA values to be returned for variants showing no association due to sampling variation in the SE. We expect variants that show association to be transformed well. For most users, i.e., those with allele frequencies, we suggest the following R syntax for OR transformation 'LmToOddsRatio(test_results, test_results$K, 0)'.

If you are unsure about the OR results generated please refer to the second paragraph of the Discussion of the above paper for some more transformations that can be used for mental calculations. Please note that the performance of these forms of the transformations deteriote quickly and are not recommended for general use.

Please download some test data to investigate the workflow of this app, which is tab delimited.


The application workflow is as follows:

i) the user loads a set of summary statistics results generated from a linear mixed model GWAS analysis using Browse button under the 'Choose file' heading. A HEADER FOR THE FILE IS ASSUMED. Different delimeters can be chosen. A sample prevalence value must be stipulated before loading. If each variant or a set of variants has a different sample prevalence then this can be supplied as a new column of the to be loaded summary statistics set. This column must be named 'K'.

ii) the user then inspects the loaded summary statistics results generated and the mapped OR in the 'Load and inspect' tab.

iii) the user can then download the results from the 'Download' tab.

The 'Individual locus test' tab can be used to map individual effects or check results from a larger set.

For further questions, please contact This app was developed by Luke Lloyd-Jones.

Browse for file to load data


Transformed odds ratio

Adjust the inputs on the left to investigate a single variant transformation.