29:null 2a:["$","title",null,{"children":"23andMe Beagle Imputation in Google Cloud | lorenasandoval88 | Observable"}] 2f:T26fb,

Imputation Summary

Workflow

63 PGP 23andMe v5 / GRCh37 files
        │
        ▼
23andMe conversion + cohort merge
        │
        ▼
22 chromosome target files
chr1–chr22 VCF/BCF
        │
        ▼
Google Cloud Storage
        │
        ▼
Google Cloud Batch
        │
        ▼
Compute Engine Linux VM
        │
        ▼
Bash job script
        │
        ├── gcloud storage
        │      └── input / output transfer
        │
        ├── target VCF
        │      └── 63 PGP participants
        │
        ├── 1000 Genomes Phase 3
        │      └── chromosome .bref3
        │
        └── GRCh37 PLINK genetic map
                │
                ▼
             Java JVM
                │
                ▼
           Beagle 5.4
                │
                ▼
    phasing + genotype imputation
                │
                ▼
        imputed chr1–chr22 .vcf.gz

Pipeline Summary

Data

Source
Personal Genome Project (PGP) ↗

Target cohort
63 participants

Genotype platform
23andMe v5

Genome build
GRCh37 / hg19

Preparation

Processing
TXT → cleaned genotypes → BCF/VCF → cohort merge → QC → chr1–22

Target files
22 indexed chromosome VCFs
chr1.vcf.gz – chr22.vcf.gz

VCF utilities
bcftools · bgzip · tabix

Reference

Reference panel
1000 Genomes Phase 3

Reference format
Beagle .bref3

Genetic maps
GRCh37 PLINK maps

Compute

Imputation software
Beagle 5.4

Runtime
Java JVM

Job orchestration
Bash + Google Cloud Batch

Compute
Compute Engine Linux VM

Storage
Google Cloud Storage

Output

Primary output
Phased + imputed chr1–chr22 .vcf.gz files

30:T197b,

Languages, Runtimes, and Tools

Component	Type	Role in Pipeline	Why It Is Used
Java (JVM)	Runtime	Executes Beagle 5.4	Runs the Beagle `.jar` with configurable memory and multithreading.
Beagle 5.4	Imputation software	Phasing + genotype imputation	Uses target genotypes and reference haplotypes to infer unobserved variants.
Bash	Shell scripting	Runs each Batch task	Coordinates file transfer, Beagle execution, and output handling.
Google Cloud Batch	Job orchestration	Schedules chromosome jobs	Provisions VMs, runs tasks, manages retries, and removes compute resources.
Compute Engine	Compute runtime	Provides Linux VM resources	Supplies the CPU, RAM, and disk required by Beagle.
Google Cloud Storage `gcloud storage`	Storage / CLI	Stores + transfers files	Stores target VCFs, reference files, genetic maps, software, logs, and outputs.
bcftools / bgzip / tabix	Genomic utilities	Processes VCF/BCF files	Filters, compresses, indexes, and validates genomic files.
1000 Genomes Phase 3 `.bref3`	Reference panel	Provides phased haplotypes	Supplies reference haplotypes used by Beagle to infer missing variants.
PLINK genetic maps	Genetic map	Provides recombination positions	Allows Beagle to model genetic distance and recombination along each chromosome.
GRCh37 / hg19	Genome assembly	Defines genomic coordinates	Keeps target genotypes, reference variants, and genetic maps aligned.
Node.js / JavaScript	Preprocessing	Prepares 23andMe data	Supports custom file selection, cleaning, conversion, and QC before imputation.

Core stack: Google Cloud Batch → Compute Engine → Bash → Java JVM → Beagle 5.4

31:T4a9,--- ## 1. What data were prepared?
The target genotype data were obtained from publicly available participant data from the Personal Genome Project (PGP). PGP participant metadata and associated genetic files were collected and consolidated to identify participants with available **23andMe genotype data**. For this imputation analysis, files were restricted to: - **23andMe v5** genotype files - **GRCh37 / Build 37** - Valid 23andMe genotype files - One selected genotype file per participant - Autosomal chromosomes **1–22** After filtering, the final cohort contained **63 participants** with 23andMe v5 / GRCh37 genotype data.
### 23andMe Data Preparation The **63 23andMe v5 / GRCh37 samples** were previously prepared in the Google Cloud pipeline documented here: [Imputation of 23andMe Files](https://observablehq.com/@lorenasandoval88/imputation-of-23andme-files)
The pipeline created indexed chromosome-level target VCFs: `chr1.vcf.gz`–`chr22.vcf.gz` These chromosome-level VCF files therefore serve as the **target genotype data for Beagle 5.4**. 32:T2eca, ### A. Create a working directory Open Google Cloud Shell: ```bash mkdir -p ~/beagle_imp cd ~/beagle_imp ``` Place both downloaded files in this directory: ```text ~/beagle_imp/ ├── run_beagle_task.sh └── beagle-batch.json ``` Confirm: ```bash ls -lh ``` --- ### B. Make the Beagle task script executable ```bash chmod +x run_beagle_task.sh ``` Confirm: ```bash ls -l run_beagle_task.sh ``` The permissions should contain `x`, for example: ```text -rwxr-xr-x ``` The `x` indicates that the shell script is executable. --- ### C. Upload the Beagle task script to Cloud Storage The Batch VMs need access to `run_beagle_task.sh`, so the script must also be copied from Cloud Shell to GCS. The Batch configuration expects the script at: ```text gs://all23_0710/beagle_imp/run_beagle_task.sh ``` Upload it: ```bash gcloud storage cp \ ~/beagle_imp/run_beagle_task.sh \ gs://all23_0710/beagle_imp/run_beagle_task.sh ``` Confirm: ```bash gcloud storage ls \ gs://all23_0710/beagle_imp/ ``` Expected: ```text gs://all23_0710/beagle_imp/run_beagle_task.sh ``` > `beagle-batch.json` does **not** need to be uploaded to GCS. > It remains in `~/beagle_imp/` and is read locally by `gcloud` when the Batch job is submitted. If `run_beagle_task.sh` is edited later in Cloud Shell, upload it again before launching a new Batch job so that the GCS copy is current. --- ### D. Download and organize required Beagle resources The Beagle workflow requires three external resources: 1. **Beagle 5.4 software** 2. **GRCh37 genetic maps** 3. **1000 Genomes Phase 3 GRCh37 Beagle reference files** These resources will be stored under: ```text gs://all23_0710/beagle/ │ ├── software/ │ └── beagle.29Oct24.c8e.jar │ ├── maps/ │ ├── plink.chr1.GRCh37.map │ ├── plink.chr2.GRCh37.map │ ├── ... │ └── plink.chr22.GRCh37.map │ └── reference/ ├── chr1.1kg.phase3.v5a.b37.bref3 ├── chr2.1kg.phase3.v5a.b37.bref3 ├── ... └── chr22.1kg.phase3.v5a.b37.bref3 ``` --- #### D1. Download Beagle 5.4 Move into the Cloud Shell working directory: ```bash cd ~/beagle_imp ``` Download the Beagle 5.4 JAR: ```bash wget \ https://faculty.washington.edu/browning/beagle/beagle.29Oct24.c8e.jar ``` Upload it to GCS: ```bash gcloud storage cp \ beagle.29Oct24.c8e.jar \ gs://all23_0710/beagle/software/ ``` Confirm: ```bash gcloud storage ls \ gs://all23_0710/beagle/software/ ``` Expected: ```text gs://all23_0710/beagle/software/beagle.29Oct24.c8e.jar ``` --- #### D2. Download the GRCh37 genetic maps Download the GRCh37 map archive: ```bash wget \ https://bochet.gcc.biostat.washington.edu/beagle/genetic_maps/plink.GRCh37.map.zip ``` Create a temporary extraction directory: ```bash mkdir -p /tmp/beagle_maps ``` Extract the maps: ```bash unzip \ plink.GRCh37.map.zip \ -d /tmp/beagle_maps ``` Check the extracted files: ```bash find /tmp/beagle_maps \ -type f \ -name "plink.chr*.GRCh37.map" ``` Expected files include: ```text plink.chr1.GRCh37.map plink.chr2.GRCh37.map ... plink.chr22.GRCh37.map ``` Upload all chromosome maps to GCS: ```bash find /tmp/beagle_maps \ -type f \ -name "plink.chr*.GRCh37.map" \ -exec gcloud storage cp {} gs://all23_0710/beagle/maps/ \; ``` Confirm: ```bash gcloud storage ls \ gs://all23_0710/beagle/maps/ ``` Expected: ```text gs://all23_0710/beagle/maps/plink.chr1.GRCh37.map gs://all23_0710/beagle/maps/plink.chr2.GRCh37.map ... gs://all23_0710/beagle/maps/plink.chr22.GRCh37.map ``` --- #### D3. Download the 1000 Genomes Phase 3 GRCh37 reference panel The Beagle-compatible 1000 Genomes Phase 3 GRCh37 reference files are stored in `bref3` format. Download chromosomes 1–22 one at a time, upload each reference file to GCS, and remove the temporary local copy: ```bash BASE="https://bochet.gcc.biostat.washington.edu/beagle/1000_Genomes_phase3_v5a/b37.bref3" for CHR in {1..22} do FILE="chr\${CHR}.1kg.phase3.v5a.b37.bref3" echo "======================================" echo "Downloading chromosome \${CHR}" echo "======================================" wget \ "\${BASE}/\${FILE}" \ -O "/tmp/\${FILE}" echo "Uploading chromosome \${CHR} to GCS" gcloud storage cp \ "/tmp/\${FILE}" \ "gs://all23_0710/beagle/reference/\${FILE}" rm "/tmp/\${FILE}" done ``` Downloading, uploading, and deleting one chromosome at a time avoids filling the Cloud Shell disk with all reference files simultaneously. Confirm: ```bash gcloud storage ls \ gs://all23_0710/beagle/reference/ ``` Expected: ```text gs://all23_0710/beagle/reference/chr1.1kg.phase3.v5a.b37.bref3 gs://all23_0710/beagle/reference/chr2.1kg.phase3.v5a.b37.bref3 ... gs://all23_0710/beagle/reference/chr22.1kg.phase3.v5a.b37.bref3 ``` --- #### D4. Confirm the Beagle resource structure Check the three Beagle resource directories: ```bash gcloud storage ls gs://all23_0710/beagle/software/ gcloud storage ls gs://all23_0710/beagle/maps/ gcloud storage ls gs://all23_0710/beagle/reference/ ``` --- #### D5. Confirm the prepared 23andMe target VCFs The 63 prepared 23andMe v5 / GRCh37 samples were previously cleaned, merged, quality controlled, and split into chromosome-level VCFs. The target files are stored under: ```text gs://all23_0710/v5_imputation_63/merge/chromosomes/ ``` Confirm: ```bash gcloud storage ls \ gs://all23_0710/v5_imputation_63/merge/chromosomes/ ``` Expected chromosome-level input files: ```text chr1.vcf.gz chr2.vcf.gz chr3.vcf.gz ... chr22.vcf.gz ``` For chromosome `N`, each Batch task therefore uses: ```text Target: gs://all23_0710/v5_imputation_63/merge/chromosomes/chrN.vcf.gz Reference: gs://all23_0710/beagle/reference/chrN.1kg.phase3.v5a.b37.bref3 Genetic map: gs://all23_0710/beagle/maps/plink.chrN.GRCh37.map Beagle: gs://all23_0710/beagle/software/beagle.29Oct24.c8e.jar ``` Conceptually: ```text 23andMe chrN.vcf.gz + 1000 Genomes Phase 3 chrN reference + GRCh37 chrN genetic map + Beagle 5.4 ↓ chrN.beagle.vcf.gz ``` --- ### E. Enable Google Cloud Batch This only needs to be done once for the Google Cloud project: ```bash gcloud services enable \ batch.googleapis.com \ compute.googleapis.com ``` --- ### F. Launch chr1–chr22 Move into the Cloud Shell working directory: ```bash cd ~/beagle_imp ``` Create a unique Batch job name: ```bash JOB="beagle-chr1-22-$(date +%Y%m%d-%H%M)" ``` Submit the Batch job using the local `beagle-batch.json` file: ```bash gcloud batch jobs submit "$JOB" \ --location=us-east4 \ --config=beagle-batch.json ``` check batch jobs ```bash gcloud batch jobs list \ --location=us-east4 ``` The Batch configuration creates **22 chromosome tasks**: ```text Task 0 → chr1 Task 1 → chr2 Task 2 → chr3 ... Task 21 → chr22 ``` The chromosome number is generated inside `run_beagle_task.sh` using: ```bash CHR=$((BATCH_TASK_INDEX + 1)) ``` Therefore: ```text BATCH_TASK_INDEX=0 → chr1 BATCH_TASK_INDEX=1 → chr2 BATCH_TASK_INDEX=2 → chr3 ... BATCH_TASK_INDEX=21 → chr22 ``` Each task independently receives: ```text chrN.vcf.gz + chrN.1kg.phase3.v5a.b37.bref3 + plink.chrN.GRCh37.map ↓ Beagle 5.4 ↓ chrN.beagle.vcf.gz ``` --- ### G. Monitor the job View the overall Batch job status: ```bash gcloud batch jobs describe "$JOB" \ --location=us-east4 ``` View the individual chromosome tasks: ```bash gcloud batch tasks list \ --job="$JOB" \ --location=us-east4 ``` Typical task states include: ```text QUEUED SCHEDULED RUNNING SUCCEEDED FAILED ``` If the Batch configuration contains: ```json "taskCount": 22, "parallelism": 4 ``` all **22 chromosomes** will be processed, but only **4 chromosome tasks will run simultaneously**. For example: ```text First group: chr1 ─┐ chr2 │ chr3 ├── running chr4 ─┘ chr5–chr22 → queued ``` As one running task finishes, another queued chromosome can begin. --- ### H. Check Beagle outputs Successful chromosome tasks write their results to: ```text gs://all23_0710/beagle_imp/output/ ``` List the results: ```bash gcloud storage ls \ gs://all23_0710/beagle_imp/output/ ``` Expected output: ```text chr1.beagle.vcf.gz chr1.beagle.vcf.gz.tbi chr1.beagle.log chr2.beagle.vcf.gz chr2.beagle.vcf.gz.tbi chr2.beagle.log ... chr22.beagle.vcf.gz chr22.beagle.vcf.gz.tbi chr22.beagle.log ``` Each chromosome therefore produces: ```text Imputed VCF VCF index Beagle log ``` The chromosome VCF contains the Beagle-imputed genotype information, including imputed genotype dosages and imputation-quality metrics. --- ### I. Workflow Summary ```text Observable notebook attachments │ ├── run_beagle_task.sh └── beagle-batch.json │ ▼ Download to Cloud Shell │ ▼ ~/beagle_imp/ ├── run_beagle_task.sh └── beagle-batch.json │ ├── beagle-batch.json │ stays in Cloud Shell │ └── run_beagle_task.sh copied to GCS │ ▼ gs://all23_0710/beagle_imp/ run_beagle_task.sh │ │ ├───────────────────────────────┐ │ │ ▼ ▼ Prepared target data Beagle resources │ │ chr1.vcf.gz ... chr22.vcf.gz Beagle 5.4 JAR │ 1000G Phase 3 reference │ GRCh37 genetic maps │ │ └───────────────┬───────────────┘ │ ▼ Google Cloud Batch │ 22 chromosome tasks │ ┌──────────────────────┼──────────────────────┐ ▼ ▼ ▼ chr1 chr2 chr3 ... chr22 │ │ │ ▼ ▼ ▼ Beagle Beagle Beagle │ │ │ └──────────────────────┼──────────────────────┘ │ ▼ chr1.beagle.vcf.gz ... chr22.beagle.vcf.gz │ ▼ Imputed genotype dosages + imputation-quality metrics ``` The same Beagle parameters, 1000 Genomes Phase 3 GRCh37 reference panel, and GRCh37 genetic maps are applied independently to each chromosome. 33:Te76,toc = { const slugify = text => text .toLowerCase() .normalize("NFKD") .replace(/[\u0300-\u036f]/g, "") .replace(/[^a-z0-9]+/g, "-") .replace(/^-|-$/g, ""); return function toc({ selector = "h2", //h2", h3", title = "Table of Contents" } = {}) { const nav = document.createElement("nav"); nav.className = "notebook-toc"; let scheduled = false; function getAllHeaders() { const cells = [ ...document.querySelectorAll(".observablehq") ]; const headers = cells.flatMap(cell => [...cell.querySelectorAll(selector)] ); // Remove duplicates and headings inside the TOC. return [...new Set(headers)] .filter(header => !nav.contains(header)) .filter(header => header.textContent.trim()) .filter(header => header.isConnected); } function render() { scheduled = false; const headers = getAllHeaders(); const usedIds = new Set( [...document.querySelectorAll("[id]")] .filter(element => !headers.includes(element)) .map(element => element.id) ); for (const header of headers) { if (!header.id) { const base = slugify(header.textContent.trim()) || "section"; let id = base; let number = 2; while (usedIds.has(id)) { id = `${base}-${number++}`; } header.id = id; } usedIds.add(header.id); } const levels = headers.map(header => Number(header.tagName.substring(1)) ); const minimumLevel = levels.length ? Math.min(...levels) : 1; nav.replaceChildren(); const heading = document.createElement("strong"); heading.textContent = title; nav.appendChild(heading); const links = document.createElement("div"); links.className = "notebook-toc-links"; if (headers.length === 0) { const emptyMessage = document.createElement("div"); emptyMessage.textContent = "No headings found."; emptyMessage.style.color = "#777"; emptyMessage.style.marginTop = "0.5rem"; links.appendChild(emptyMessage); } for (const header of headers) { const level = Number( header.tagName.substring(1) ); const link = document.createElement("a"); link.href = `#${header.id}`; link.textContent = header.textContent.trim(); link.dataset.level = level; link.style.display = "block"; link.style.marginLeft = `${(level - minimumLevel) * 1.25}rem`; link.style.padding = "0.2rem 0"; link.style.textDecoration = "none"; link.onclick = event => { event.preventDefault(); header.scrollIntoView({ behavior: "smooth", block: "start" }); history.replaceState( null, "", `#${header.id}` ); }; links.appendChild(link); } nav.appendChild(links); } function scheduleRender() { if (scheduled) return; scheduled = true; requestAnimationFrame(render); } render(); const observer = new MutationObserver(mutations => { const onlyTocChanges = mutations.every(mutation => nav.contains(mutation.target) ); if (!onlyTocChanges) { scheduleRender(); } }); observer.observe(document.body, { childList: true, subtree: true, characterData: true }); invalidation.then(() => { observer.disconnect(); }); return nav; }; }34:[] 2d:{"id":"b637a2da98fce939","slug":"23andme-imputation-pipeline-beagle","title":"23andMe Beagle Imputation in Google Cloud","thumbnail":"00e5c6d89a954372ddfaf81bca30716258ab60903532c69455c7360d05fedafd","access_level":"public","create_time":"$D2026-07-21T14:22:24.574Z","update_time":"$D2026-09-21T14:17:10.009Z","view_time":null,"first_public_time":"$D2026-07-21T17:06:40.667Z","has_unpublished_edits":null,"fork":false,"stars":1,"starred":false,"owner":{"type":"team","id":"5a5147324ed15c62","login":"lorenasandoval88","name":"","avatar_url":"https://avatars-next.observableusercontent.com/c8926e2434569cc1f72aeac62929e824fdec3f0ec037d4a41a38a7728d14e950","tier":"starter_2024"},"version":465,"publish_version":null,"role":null,"authored":false,"authors":[{"type":"individual","id":"5c479a2c5ae5f52f","login":"lorenasandoval","name":"","avatar_url":"https://avatars-next.observableusercontent.com/c8926e2434569cc1f72aeac62929e824fdec3f0ec037d4a41a38a7728d14e950","tier":null}],"members":[{"type":"individual","id":"5c479a2c5ae5f52f","login":"lorenasandoval","name":"","avatar_url":"https://avatars-next.observableusercontent.com/c8926e2434569cc1f72aeac62929e824fdec3f0ec037d4a41a38a7728d14e950","tier":null,"role":"editor"},{"type":"team","id":"5a5147324ed15c62","login":"lorenasandoval88","name":"","avatar_url":"https://avatars-next.observableusercontent.com/c8926e2434569cc1f72aeac62929e824fdec3f0ec037d4a41a38a7728d14e950","tier":"starter_2024","role":"editor"}],"comments":[],"custom_thumbnail":false,"trashed":false,"lapsed":null,"subscribed":null,"forks":2,"collection":{"id":"8bdfe82d7003e3b0","slug":"imputation","title":"imputation","owner":{"id":"5a5147324ed15c62","login":"lorenasandoval88","name":""}},"type":"notebook","body":{"title":"23andMe Beagle Imputation in Google Cloud","cells":[{"id":45,"value":"# 23andMe Beagle Imputation in Google Cloud (batch job)\n","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":115,"value":"toc()","pinned":true,"hidden":false,"mode":"ts","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":59,"value":"FileAttachment(\"image (2).png\").image()","pinned":false,"hidden":false,"mode":"ts","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":105,"value":"FileAttachment(\"image.png\").image()","pinned":false,"hidden":false,"mode":"ts","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":110,"value":"---","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":162,"value":"$2f","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":159,"value":"$30","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":149,"value":"","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":61,"value":"$31","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":0,"value":"\n---\n\n## 2. What will Beagle use?\n
\n### Reference-Based Imputation with Beagle\n\nThe prepared 23andMe target data will next be imputed using:\n\n- **Beagle 5.4**\n- **1000 Genomes Phase 3 reference panel**\n- **GRCh37**\n- **2,504 reference individuals**\n- **5,008 phased reference haplotypes**\n- **GRCh37 recombination maps**\n\nBeagle uses the prepared 23andMe VCFs as the target data, the 1000 Genomes Phase 3 panel as the phased haplotype reference, and GRCh37 genetic maps to model recombination.\n\n","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"","database":"$undefined"},{"id":66,"value":"---\n\n## 3. Where are the files stored?\n
\n### Cloud bucket and folder structure\n\nThe local Cloud Shell directory contains the launch files, while the GCS bucket stores the target data, reference resources, and persistent Beagle outputs.\n```\nCloud Shell\n~/beagle_imp/\n│\n├── run_beagle_task.sh\n└── beagle-batch.json\n\n\nGCS\ngs://all23_0710/\n│\n├── v5_imputation_63/merge\n│ └── chromosomes/\n│ ├── chr1.vcf.gz\n│ ├── chr2.vcf.gz\n│ └── ...\n│\n├── beagle/\n│ ├── software/\n│ │ └── beagle.29Oct24.c8e.jar\n│ ├── reference/\n│ │ ├── chr1.1kg.phase3.v5a.b37.bref3\n│ │ └── ...\n│ └── maps/\n│ ├── plink.chr1.GRCh37.map\n│ └── ...\n│\n└── beagle_imp/\n ├── run_beagle_task.sh\n └── output/\n ├── chr1.beagle.vcf.gz\n ├── chr1.beagle.vcf.gz.tbi\n ├── chr1.beagle.log\n ├── ...\n └── chr22.beagle.log\n ```","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":97,"value":"---\n## 4. How do I actually run it?\n\n
\n\n### Run Beagle chr1–chr22 in Google Cloud Batch\n\nThe two files used to launch the Beagle Batch workflow are:\n\n```text\nrun_beagle_task.sh\nbeagle-batch.json\n```\n\nThese files are stored as attachments with this Observable notebook for reproducibility.\n\nFor this workflow:\n\n1. **`beagle-batch.json` only needs to be in Cloud Shell.**\n It is read locally when the Batch job is submitted.\n\n2. **`run_beagle_task.sh` is kept in Cloud Shell and copied to Cloud Storage.**\n The Cloud Shell copy is the editable working copy, while the Batch VMs execute the copy stored in GCS.\n\nThe resulting organization is:\n\n```text\nCloud Shell\n~/beagle_imp/\n├── run_beagle_task.sh\n└── beagle-batch.json\n\nCloud Storage\ngs://all23_0710/beagle_imp/\n└── run_beagle_task.sh\n```\n\nTo execute the workflow, download both attached files to **Google Cloud Shell** and run the following steps.\n\n---","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":72,"value":"$32","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":113,"value":"","pinned":true,"hidden":false,"mode":"ts","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":5,"value":"---\n## Primary Technical References\n\n- [Minimac4 official repository](https://github.com/statgen/minimac4)\n- [Eagle v2.4.1 documentation](https://alkesgroup.broadinstitute.org/Eagle/)\n- [Beagle 5.5 official page](https://faculty.washington.edu/browning/beagle/beagle.html)\n- [Beagle 5.5 documentation](https://faculty.washington.edu/browning/beagle/beagle_5.5_17Dec24.pdf)\n- [conform-gt documentation](https://faculty.washington.edu/browning/conform-gt.html)\n- [FastImpute paper](https://openbioinformaticsjournal.com/VOLUME/19/ELOCATOR/e18750362421210/FULLTEXT/)","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"","database":"$undefined"},{"id":24,"value":"---\n## Helper Functions","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"","database":"$undefined"},{"id":26,"value":"$33","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"}],"files":"$W34","stdlib":"1","titling":"auto"}} 2e:"$2d:body"