UK Biobank Research Application Platform (UKB RAP) Educational Materials

The docker_apps directory contains resources for reproducible research using Docker within the DNAnexus environment. It includes basic scripts for managing Docker images and example applications that utilize Docker snapshot files.

Key components include:

• Docker Image Management: Instructions for pulling and saving Docker snapshot images to a project.
• Swiss Army Knife Integration: Basics for running Docker images using Swiss Army Knife.
• Example Applications: A sample application (samtools_count_docker) demonstrating how to use a Docker snapshot file.

This repository contains supplementary materials related to two specific resources:

1. DNAnexus Blog Post: Discusses infrastructure challenges for machine learning in the cloud using UK Biobank data. Link to Blog Post
2. Seminar "Jak stary je vas mozek": Materials for the seminar held at the Faculty of Mathematics and Physics, Prague (28 Feb 2024). Information can be found at the MFF Seminar website or via the seminar poster.

This workflow converts UK Biobank (UKB) array genotyping data from the GRCh37 reference build to the GRCh38 (specifically hs38DH) assembly. This is necessary for performing association testing (Step 2 of GWAS) using sequencing data (WES/WGS) which is released on GRCh38.

The pipeline uses Picard LiftoverVcf to handle scenarios where an ALT allele in the source build equals the target's REF allele by swapping REF/ALT variants. To optimize memory, the pipeline disables variant sorting within Picard LiftoverVcf and instead uses bcftools sort on intermediate VCFs.

Key Resource Requirements:

• Disk Space: Up to 200 GB per chromosome (approximately 30x the original BED file size).
• Memory: Approximately 50 GB per chromosome.
• Tools used: bcftools, picard, plink, and plink2.

The proteomics directory contains two primary analysis examples:

• protein_DE_analysis: Contains JupyterLab notebooks for performing differential expression analysis on UKB-RAP data.
• protein_pQTL: Contains a notebook designed to prepare data for input into the REGENIE app on the UKB-RAP.

This repository serves as a collection of tutorials and webinar materials for the UK Biobank Research Application Platform (UKB RAP).

To get started with the platform or find community support, use the following resources:

• Platform Documentation/Access: https://ukbiobank.dnanexus.com/
• Community Support & Issue Reporting: Use the DNAnexus Community Forum to ask questions or submit issues regarding the code found in this repository.

The REGENIE app automatically selects an instance type based on an internal formula. You can use this formula to approximate the RAM required for your analysis to estimate costs and choose appropriate instance types.

RAM (Storage in MB) Formula: RAM = (#samples * (#vas / block_size) / 1048576) * 40 * #traits + 7000

Example benchmark: Running REGENIE on 1 chromosome (chr22) with 200 protein traits took ~19.5 hours using:

• Step 1: mem3_ssd2_v2_x4 (4500 MB / 30000 MB)
• Step 2: mem1_ssd1_v2_x2 (800 MB / 47000 MB)

You can use regular expressions to generate a TSV (Tab-Separated Values) file where each line represents a single job. This file can then be passed to the dx run command to submit all jobs in the batch.

Command:

dx run --batch-tsv TSV_file.txt

The samtools_count_apt applet uses samtools to count alignments from a CRAM file and outputs the count into a separate TXT file. You can execute this app on the DNAnexus Platform using the dx run command.

To run the app, provide the path to your CRAM file using the -imappings_cram input flag and use the -y flag to run it immediately.

dx run samtools_count_apt -imappings_cram=<path_to_your_cram_file> -y

To convert BGEN files using the web interface:

1. Open the Tool library.
2. Select Swiss-Army-Knife.
3. On the Swiss-Army-Knife home page, click the Versions tab and select version 4.1.1.
4. Click Run.
5. Select your project and the required files in the pop-up menu.
6. In the Analysis Inputs 1 tab, set the following:
   • Input files: <filename>.bgen and <filename>.sample.
   • Command line: qctool -g *.bgen -s *.sample -og <filename>_converted.bgen -os <filename>_converted.sample -ofiletype bgen -bgen-bits 8 -bgen-compression zlib -bgen-omit-sample-identifier-block