Data release overview

Background

This dataset contains a set of association test summary statistics for association tests between severe malaria cases and population controls collected in eleven populations as shown in the following table:

All cases were diagnosed with meeting WHO definitions of severe malaria (see reference [2]), while controls were samples from within the general population and from new births. Underlying genotypes for these samples have also been deposited in the European Genome-Phenome Archive (EGA) under study ID EGAS00001001311 https://ega-archive.org/studies/EGAS00001001311.

This README file contains information on on the following relevant to the dataset:

• Information on terms of use and citation information for this data
• A summary of methods used to create this dataset
• Detailed information on the data contents

Terms of use

The summary statistics included here, including allele frequency estimates and association test effect size estimates, parameter standard errors and evidence measures, are made freely available for use. Please include the following text as an acknowledgement in any publication arising from the use of this data: “This study makes use of data generated by MalariaGEN. A full list of the investigators who contributed to the generation of the data is available from www.malariagen.net. Funding for this project was provided by Wellcome Trust (WT077383/Z/05/Z) and the Bill & Melinda Gates Foundation through the Foundation of the National Institutes of Health (566) as part of the Grand Challenges in Global Health Initiative.”

References and citation

This dataset was created for and described in the following manuscript:

[1] Malaria Genomic Epidemiology Network, “New insights into malaria susceptibility from the genomes of 17,000 individuals from Africa, Asia, and Oceania”, Nature Communications (2019). https://doi.org/10.1038/s41467-019-13480-z; bioRxiv link: https://doi.org/10.1101/535898

Please cite the above article if you make use of data from this release in disseminated work.

Details of data

This release includes summary statistics from frequentist and Bayesian meta-analysis of association test with severe malaria (SM), as well as severe malaria subphenotypes (CM, SMA and OTHER) as defined in the following table:

The following table summarises the available files:

Below we describe the contents of each of the files provided in this dataset.

Methods summary

A full set of methods can be found in [1]. In brief, genotypes were obtained by typing each sample on the Illumina Omni 2.5M platform, followed by imputation into the 1000 Genomes Reference panel (1000GP) and into a custom panel (“combined panel”) obtained by adding additional whole-genome sequenced samples (https://ega-archive.org/studies/EGAS00001003648) to the 1000GP. Additionally, HLA alleles were imputed using HLA*IMP:02 and glycophorin region CNV alleles were imputed using a panel described previously (Leffler et al, https://doi.org/10.1126/science.aam6393).

Association tests were conducted by logistic regression in each of the eleven populations in Table 1, include 5 principal components as covariates, using SNPTEST (http://www.well.ox.ac.uk/~gav/snptest). Association test results were then meta-analysed using BINGWA (http://www.well.ox.ac.uk/~gav/bingwa), under both frequentist fixed-effect meta-analysis and a flexible bayesian meta-analysis framework described in [1].

Additionally, we implemented a multinomial logistic regression method to test each genetic variant against severe malaria subtype as defined in our data (CM, SMA or other severe malaria; c.f. table 2). Case samples identified as having both CM and SMA were excluded from these tests. Subphenotype association test results were also meta-analysed using BINGWA in a multivariate meta-analysis framework.

Common columns

The meta-analysis results files have the following common columns:

Note: In all files the variant_id column refers to the specific variant (i.e. the specific combination of chromosome, position and alleles) being analysed. The analysis_id and analysis columns refer to the imputation reference panel from which the variant was imputed. To match results between files, users should match on both the variant_id and analysis_id columns. Imputation refrence panels are detialed in the following table.

Combined meta-analysis results

An overview of meta-analysis results can be found in this file:

MalariaGEN_summary_statistics_combined_evidence.csv.gz

This is a gzipped comma-seperated file which contains overall measures of evidence under additive, dominant, recessive and heterozygote modes of inheritance, as well as an overall model-averaged Bayes factor and indication of the best-fitting model.

This file has the following columns:

Further notes:

• The effective_minor_allele_count (EMAC) column is computed as the sum over populations of the minor allele count times the IMPUTE info score. Only variants with EMAC >= 250 are included in this release.

• The values in the included_cohorts are strings of eleven 0’s and 1’s. These correspond to the eleven populations in the (roughly west-east) order in Table 1. A 1 indicates that the effect size estimate (for additive, case-control association model) for this population was included in meta-analysis. A 0 indicates that it was excluded due to low per-population minor allele count, low per-population IMPUTE info score, or failure to fit the model.

• Model-averaged Bayes factors were computed under the set of prior weights detailed in [1].

Mode-specific frequentist meta-analysis results

Detailed meta-analysis results for fixed-effect meta-analysis under specific mode of inheritance are available in these files:

`MalariaGEN_case-control:[mode]:effects.csv.gz`
`MalariaGEN_subphenotype:[mode]:effects.csv.gz`

where mode is ‘add’ (for additive model), ‘dom’ (dominant effect of the non-reference allele), ‘rec’ (recessive effect of the non-reference allele), or ‘het’ (heterozygote effect). The case-control files contain results of meta-analysis of logistic regression against case-control status in each population, and the subphenotype files contain results of meta-analysis analysis of multinomial logistic regression against malaria subtypes in each population.

The following tables list the columns of these files.

Detailed bayesian meta-analysis results

Detailed meta-analysis results for bayesian meta-analysis under specific mode of inheritance are available in these files:

`MalariaGEN_case-control:[mode]:bfs.csv.gz`
`MalariaGEN_subphenotype:[mode]:bfs.csv.gz`

where [mode] is one of: add (additive effect), dom (dominant effect of the non-reference allele), rec (recessive effect of the non-reference allele), or het (heterozygote effect).

Results are presented as a set of Bayes factors (BFs). All Bayes factors were computed assuming an asymptotic approximation and a Gaussian prior on the effect size variance σ², and we used an equal mixture of σ=0.2, 0.4, 0.6, 0.8 throughout. Details of model assumptions and prior weights can be found in [1].

The following table lists the columns of the case-control Bayesian analysis files.

Bayes factor columns for population groups are encoded using the populations indicator, which is a string of 0’s and 1’s indicating whether the effect is assumed nonzero or zero in each population. For this purpose are taken in the order shown in Table 1 (i.e. roughly west-east order). Examples are given below:

See [1] for full details of models included.

The following table lists the columns of the subphenotype Bayesian analysis files.

Per-population results files

Per-population results, including estimated allele frequency estimates, IMPUTE info scores, and per-population association test results can be found in these files:

`MalariaGEN_case-control:[mode]:percohort.csv.gz`
`MalariaGEN_subphenotype:[mode]:percohort.csv.gz`

The following table lists columns common to these files in addition to those listed above:

In the above, [population] refers to the acronym column in Table 1, i.e. is one of Gambia, Mali, BurkinaFaso, Ghana, Nigeria, Cameroon, Malawi, Tanzania, Kenya, Vietnam, or PNG. Results are provided for all eleven populations.

The following table lists association test-related columns found in the case-control files:

The following table lists association test-related columns found in the subphenotype files:

URLs

• SNPTEST - http://www.well.ox.ac.uk/~gav/qctool
• BINGWA - http://www.bgenformat.org