Doing Genomics in the Cloud: From RNA to Result

Computational genomics has become a core toolkit in the study of biological systems at the molecular level. To run genomics workflows, a researcher needs access to advanced computer systems including compute, storage, memory, and networks to move and mine huge genomics datasets. The workflows will include pulling high-throughput DNA datasets from the NCBI-SRA data repository, performing reference genome mapping of SRA RNAseq datasets, and building a gene co-expression network.

The Workshop:

We will cover the complete deployment life cycle that enables scientific workflows to be run in the cloud. This includes:

Accessing a Kubernetes(K8s) cluster via the command line.
Creating a persistent NFS server to store workflow data, then loading workflow data onto it.
Pulling genomic RNA data from the NCBI's SRA database.
Deploying GEMmaker on the K8s cluster to create a Gene Expression Matrix.
Using the output from the GEMmaker run with Knowledge Independent Network Construction(KINC), a genomic workflow, to the K8s cloud.
Downloading the resulting Gene Coexpression Network(GCN) from the Kubernetes cluster, then visualizing the network.

0. Prerequisites

The following software is necessary to participate in the demo:

kubectl - Kubernetes CLI
Nextflow - Workflow Manager
Java
Helm(optional) - Kubernetes Deployment Orchestrator

To streamline the workshop, all software has been packaged into a virtual machine that has been replicated for each user except for Helm.

Install Helm

mkdir -p ~/Desktop/classroom/myfiles && cd ~/Desktop/classroom/myfiles

wget https://get.helm.sh/helm-v3.6.0-linux-amd64.tar.gz

tar -xvf helm-v3.6.0-linux-amd64.tar.gz

sudo cp linux-amd64/helm /usr/local/bin

Add the stable repo:

helm repo add stable https://charts.helm.sh/stable

Update Helm's repositories(similar to apt-get update):

helm repo update

An additional requirement is access to the Kubernetes clusters that will be used for the workshop.

If you do not have your CCP cluster credentials(kubeconfig) and access to your personal VM, please let us know.

Access Praxis(optional)

Navigate to the Praxis portal

Enter your credentials.

Select the class Running Scientific Workflows on Regional R&E Kubernetes Clusters Workshop

Select Learning at the upper right side of the menu bar.

Select the lab session Accessing the Cloud through c-Light CCP/IKS Cluster, when prompted start the live lab.

Once the Jupyter notebook is provisioned, select Terminal from the menu to access a Bash terminal from within your VM!

Finally, please clone the workshop repo and the dtp-personal repositories to a folder with persistent storage:

git clone https://github.com/SciDAS/scidas-workshop ~/Desktop/classroom/myfiles/scidas-workshop

git clone https://github.com/SciDAS/dtp-personal.git ~/Desktop/classroom/myfiles/dtp-personal

Access Nautilus(optional)

Nautilus is the Kubernetes cluster that composes the National Research Platform. With nodes spread across the United States and a few international sites, Nautilus is an extremely useful resource for computational scientists. To join Nautilus:

Create an account by logging in.

You may use an affiliated campus email, or a Google email.

Next, send me an email at [email protected], and I will add you to our namespace used for education. You may also request one for your own research!

Finally, download your kubeconfig by selecting "Get Config" from the home page.

Creating a Research Environment in Kubernetes

This section will cover accessing a Kubernetes cluster and creating persistent storage.

1. Access Kubernertes Cluster

Download or copy/paste the kubeconfig you were provided to a file named config.

Create a folder ~/.kube:

mkdir -p ~/.kube

Move the kubeconfig to your .kube folder:

mv config ~/.kube

Set permissions:

chmod 600 ~/.kube/config

Confirm your cluster name:

kubectl config current-context

The output should match the name of your cluster.

You now have access to your K8s cluster!

Issue an API call to view current pods(containers) that are deployed within the namespace:

kubectl get pods

If you are on Nautilus/PRP, be sure to switch to the teach-compbio namespace if you have not already(optional):

kubectl config set-context --current --namespace=teach-compbio

2. Create Persistant Data Storage to Host Workflow Data

Now it is time to provision a NFS server to store workflow data. We will streamline this process by using Helm. Helm is a Kubernetes package manager!

First, check to see if a storage class already exists:

kubectl get sc

If not, follow the instructions below in 2a.

If a valid storage class exists, proceed to 2b.

2a. Create a NFS Storage Class(optional)

On some clusters, you may not have permission to do this, or you may not need to.

Follow the instructions above to install Helm.

Install a NFS provisioner onto the K8s cluster to permit dynamic provisoning for ~300Gb of persistent data:

helm install kf stable/nfs-server-provisioner \
--set=persistence.enabled=true,persistence.storageClass=standard,persistence.size=320Gi

2b. Create a Persistent Volume Claim(PVC)

Check that the a valid storage class exists:

kubectl get sc

Next, deploy a 300Gb Persistant Volume Claim(PVC) to the cluster:

cd ~/Desktop/classroom/myfiles/scidas-workshop

Edit the file with nano pvc.yaml and enter your name and a valid storage class for your PVC!

metadata:
  name: pvc-<YOUR_NAME>
spec:
  storageClassName: rook-cephfs

kubectl create -f pvc.yaml

Check that the PVC was deployed successfully:

kubectl get pvc

3. Start a Data Transfer Pod

To interact with data on the cluster, we will use the Data Transfer Pod(DTP) Utility

The Data Tranfer Pod Utility is a tool developed to make it easy to move data in and out of a Kubernetes cluster, using a variety of protocols.

Right now, the supported protocols are:

Google Cloud SDK
Globus Connect Personal
iRODS
Named-Data Networking(NDN)
Aspera CLI
Amazon Web Services
MinIO
NCBI's SRA Tools
Fast Data Transfer(FDT)
Local transfers(to/from the user's local machine)

To start a DTP mounted to the new PVC:

Go to the repo:

cd ~/Desktop/classroom/myfiles/dtp-personal

Edit the file helm/values.yaml to set the correct PVC and enable the SRA Tools container:

nano helm/values.yaml

ExistingPVC:
  # If true, use existing PVC on local cluster.
  Enabled: true
  Name: pvc-<YOUR_NAME>
  Path: /workspace

# SRA Tools
SRATools: 
  Enabled: true

Start DTP-Personal:

./start dtp-<YOUR_NAME>

Once started, interact with the running DTP:

./interact dtp-<YOUR_NAME>

Choose 1 to enter the base container.

This tab is now on your cluster's persistent filesystem.

Create a new terminal tab with File -> New.

4. Grant Nextflow Permissions(optional)

In the new tab on your local machine's filesystem....

Give Nextflow the necessary permissions to deploy jobs to your K8s cluster:

This command only needs to be run once, and is not necessary on namespaces where Nextflow has already been run.

kubectl create rolebinding default-edit --clusterrole=edit --serviceaccount=default:default 
kubectl create rolebinding default-view --clusterrole=view --serviceaccount=default:default

GEMmaker

GEMmaker is a genomic workflow that takes raw RNA sequences and builds a Gene Expression Matrix(GEM), a matrix that compares gene expression across a number of samples. GEMs are used as input by a number of downstream workflows, including Knowledge Independent Network Construction(KINC).

1. Index Arabidopsis Reference Genome(optional)

GEMmaker needs a reference genome to map gene expression levels. There is a simple "Cool Organism"(CORG) reference genome that has been pre-indexed for each GEMmaker pipeline option(Kallisto, Salmon, Hisat2).

CORG is used by default in the configuation files, but to get biologically accurate results the real Arabidopsis Thaliana Reference Genome should be indexed.

Here are the steps to download and index the Arabidopsis genome using Kallisto:

On your local VM....

Go to the repo:

cd ~/Desktop/classroom/myfiles/scidas-workshop/gemmaker

Edit the file kallisto.yaml:

metadata:
  name: kallisto-<YOUR_NAME>
  labels:
    app: kallisto-<YOUR_NAME>
spec:
  containers:
  - name: kallisto-<YOUR_NAME>

      persistentVolumeClaim:
        claimName: pvc-<YOUR_NAME> # Enter valid PVC

Deploy the GEMMaker container to index the genome:

kubectl create -f kallisto.yaml

Access the cluster using your kallisto container:

kubectl exec -ti kallisto-<YOUR_NAME> -- /bin/bash

On the cluster....

Navigate to your input directory and download the Arabidopsis genome for indexing:

mkdir -p /workspace/gemmaker && cd /workspace/gemmaker && wget ftp://ftp.ensemblgenomes.org/pub/plants/release-50/fasta/arabidopsis_thaliana/cdna/Arabidopsis_thaliana.TAIR10.cdna.all.fa.gz

Index the Genome:

kallisto index -i /workspace/gemmaker/Arabidopsis_thaliana.TAIR10.kallisto.indexed Arabidopsis_thaliana.TAIR10.cdna.all.fa.gz

2. Pull RNA Data from NCBI

Moving data into your K8s cluster is very important. We will cover 4 methods.

All use the following list of SRA IDs:

On the cluster....

Create a folder for your workflow and input:

mkdir -p /workspace/gemmaker/input && cd /workspace/gemmaker/input

Make a file in the same folder called SRAs.txt with the SRA IDs of 26 Arabidopsis samples:

cat > /workspace/gemmaker/SRA_IDs.txt << EOL
SRR1058270
SRR1058271
SRR1058272
SRR1058273
SRR1058274
SRR1058275
SRR1058276
SRR1058277
SRR1058278
SRR1058279
SRR1058280
SRR1058281
SRR1058282
SRR1058283
SRR1058284
SRR1058285
SRR1058286
SRR1058287
SRR1058288
SRR1058289
SRR1058290
SRR1058291
SRR1058292
SRR1058293
SRR1058294
SRR1058295
EOL

Make sure it is formatted correctly!

# cat /workspace/gemmaker/SRA_IDs.txt
SRR1058270
SRR1058271
SRR1058272
SRR1058273
SRR1058274
SRR1058275
SRR1058276
SRR1058277
SRR1058278
SRR1058279
SRR1058280
SRR1058281
SRR1058282
SRR1058283
SRR1058284
SRR1058285
SRR1058286
SRR1058287
SRR1058288
SRR1058289
SRR1058290
SRR1058291
SRR1058292
SRR1058293
SRR1058294
SRR1058295

2a. Use a Built-In Workflow Utility

Many workflows, such as GEMmaker, have a built in utility to pull input data. This utility requires a path on to a file on the cluster containing a list of SRA IDs to be pulled.

2b. Use a Single-Container Deployment to Pull Files Sequentially(optional)

On your local VM....

Go to the repo:

cd ~/Desktop/classroom/myfiles/scidas-workshop/gemmaker

Edit the file sra-tools.yaml:

metadata:
  name: sra-tools-<YOUR_NAME>
  labels:
    app: sra-tools-<YOUR_NAME>
spec:
  containers:
  - name: sra-tools-<YOUR_NAME>

      persistentVolumeClaim:
        claimName: pvc-<YOUR_NAME> # Enter valid PVC

Deploy the sra-tools container:

kubectl create -f sra-tools.yaml

Check the status of your pod:

kubectl get pods

Get a Bash session inside your pod:

kubectl exec -ti sra-tools-<YOUR_NAME> -- /bin/bash

Make an input folder and enter /workspace/gemmaker:

mkdir -p /workspace/gemmaker/input && cd /workspace/gemmaker/

Initialize SRA-Tools:

printf '/LIBS/GUID = "%s"\n' 'uuidgen' > /root/.ncbi/user-settings.mkfg

Then, pull the data using the list of SRA IDs:

while read id; do prefetch "$id" && fasterq-dump "$id"/"$id".sra --split-files -O /workspace/gemmaker/input/ --force; done < /workspace/gemmaker/SRA_IDs.txt

2c. Use a Multi-Container StatefulSet to Pull Files in Parallel(optional)

A StatefulSet is an ordered deployment of containers.

To parallelize pulling data, we can create a StatefulSet with one container for each file we want to pull.

On the cluster....

Make an input folder and enter /workspace/gemmaker:

mkdir -p /workspace/gemmaker/input && cd /workspace/gemmaker/

pull-sample-batch.sh is a script that gets the ordered index of a container and pulls the SRA ID at that line of the list.

Download the script to the cluster:

cd /workspace/gemmaker/ && wget https://raw.githubusercontent.com/SciDAS/scidas-workshop/master/gemmaker/pull_sample.sh

On your local VM....

Go to the repo:

cd ~/Desktop/classroom/myfiles/scidas-workshop/gemmaker

Edit the file statefulset.yaml:

metadata:
  name: sra-batch-<YOUR_NAME>
  labels:
    app: sra-batch-<YOUR_NAME>
spec:
  serviceName: sra-batch-<YOUR_NAME>
  replicas: 7
  selector:
    matchLabels:
      app: sra-batch-<YOUR_NAME>
  template:
    metadata:
      labels:
        app: sra-batch-<YOUR_NAME>
    spec:
      containers:               
      - name: sra-batch-<YOUR_NAME>
        image: ncbi/sra-tools
        command: [ "/bin/sh", "-c", "--" ]
        args: [ "cd" ]
        resources:
          requests:
            cpu: 1
            memory: 4Gi
          limits:
            cpu: 1
            memory: 4Gi
        volumeMounts:
        - name: sra-batch-pvc
          mountPath: /workspace
      restartPolicy: Always
      volumes:
        - name: sra-batch-pvc
          persistentVolumeClaim:
            claimName: pvc-<YOUR_NAME> # Enter valid PVC

      persistentVolumeClaim:
        claimName: pvc-<YOUR_NAME> # Enter valid PVC

2d. Use the Data Transfer Pod Utility to Pull Files Sequentially(optional)

The Data Tranfer Pod Utility is a tool developed to make it easy to move data in and out of a Kubernetes cluster, using a variety of protocols.

Right now, the supported protocols are:

Google Cloud SDK
Globus Connect Personal
iRODS
Named-Data Networking(NDN)
Aspera CLI
Amazon Web Services
MinIO
NCBI's SRA Tools
Fast Data Transfer(FDT)
Local transfers(to/from the user's local machine)

The steps to pull the RNA sequences are essentially the same as in 1b once the SRA Tools container is deployed. To deploy an instance of DTP-Personal, clone the repository and follow the instructions above.

Interact with the running DTP:

./interact dtp-<YOUR_NAME>

Choose 2 to enter the sra-tools container.

The steps to pull the RNA sequences are essentially the same as in 1b once the SRA Tools container is accessed.

4. Configure/Deploy GEMmaker

On your local VM....

Go to the repo:

cd ~/Desktop/classroom/myfiles/scidas-workshop/gemmaker

If you followed step 1. and manually indexed the A. Thal. genome:

For a remote run:

nextflow -C nextflow.config kuberun systemsgenetics/gemmaker \
   -profile k8s \
   -v pvc-<YOUR_NAME> \
  --pipeline kallisto \
  --kallisto_index_path /workspace/gemmaker/Arabidopsis_thaliana.TAIR10.kallisto.indexed \
  --sras  /workspace/gemmaker/input/SRA_IDs.txt \
  --outdir /workspace/gemmaker/output \
  --max_cpus 4

For a local run:

nextflow -C nextflow.config kuberun systemsgenetics/gemmaker \
   -profile k8s \
   -v pvc-<YOUR_NAME> \
  --pipeline kallisto \
  --kallisto_index_path /workspace/gemmaker/Arabidopsis_thaliana.TAIR10.kallisto.indexed \
  --input  /workspace/gemmaker/input/*.fastq \
  --outdir /workspace/gemmaker/output \
  --max_cpus 4

If you did not, run with the CORG Genome

nextflow -C nextflow.config kuberun systemsgenetics/gemmaker \
   -profile k8s \
   -v pvc-<YOUR_NAME> \
  --pipeline kallisto \
  --kallisto_index_path /workspace/projects/systemsgenetics/gemmaker/assets/demo/references/CORG.transcripts.Kallisto.indexed \
  --sras  /workspace/gemmaker/input/SRA_IDs.txt \
  --outdir /workspace/gemmaker/output \
  --max_cpus 4

For a local run:

nextflow -C nextflow.config kuberun systemsgenetics/gemmaker \
   -profile k8s \
   -v pvc-<YOUR_NAME> \
  --pipeline kallisto \
  --kallisto_index_path /workspace/projects/systemsgenetics/gemmaker/assets/demo/references/CORG.transcripts.Kallisto.indexed \
  --input  /workspace/gemmaker/input/*.fastq \
  --outdir /workspace/gemmaker/output \
  --max_cpus 4

5. View Output

After the workflow has completed, switch tabs to your cluster's filesystem

To view the resulting GEM:

cat /workspace/gemmaker/output/GEMs/GEMmaker.GEM.TPM.txt

Copy the GEM to the KINC input folder for the next workflow(optional):

mkdir -p /workspace/kinc/input && cp /workspace/gemmaker/output/GEMs/GEMmaker.GEM.TPM.txt /workspace/kinc/input/Arabidopsis.emx.txt

Follow the next part to create and visualize a Gene Co-expression Network(GCN) from the GEM!

KINC

Knowledge Independent Network Construction is a genomic workflow that takes a Gene Expression Matrix(GEM) and generates a Gene Co-Expression Network(GCN). GCNs can be visualized and compared to discover novel gene interactions.

1. Configure KINC

Interact with the running DTP:

./interact dtp-<YOUR_NAME>

On the cluster....

Make an input folder /workspace/kinc:

mkdir -p /workspace/kinc

On your local VM....

Go to the repo:

cd ~/Desktop/classroom/myfiles/scidas-workshop/kinc

Edit the file nextflow.config if needed:

params {
    input {
        dir = "/workspace/kinc/input"
        emx_txt_files = "*.emx.txt"
        emx_files = "*.emx"
        ccm_files = "*.ccm"
        cmx_files = "*.cmx"
    }

    output {
        dir = "/workspace/kinc/output"
    }

1a. Load the Input Dataset(optional)

If you do not have an Arabidopsis GEM, you need to do this.

There are 3 datasets/GEMs in the scidas-workshop kinc directory.

Yeast
Cervix
Rice

On your local VM....

Go to the repo:

cd ~/Desktop/classroom/myfiles/scidas-workshop/kinc

Choose which one you want by copying the associated folder to input:

cp -r input-yeast/ input cp -r input-cervix/ input cp -r input-rice/ input

Use any already running pod mounted to your PVC to load the input data onto the PVC:

kubectl exec <POD_NAME> -- bash -c "mkdir -p /workspace/kinc"

kubectl cp "input" "<POD_NAME>:/workspace/kinc"

2. Deploy KINC

Deploy KINC using nextflow-kuberun:

nextflow -C nextflow.config kuberun systemsgenetics/kinc-nf -v pvc-<YOUR_NAME>

3. Retrieve and Visualize Gene Co-expression Network

Copy the output of KINC from the PVC to your VM:

cd ~/Desktop/classroom/myfiles/scidas-workshop

kubectl exec <POD_NAME> -- bash -c \
"for f in \$(find /workspace/kinc/output/ -type l); do cp --remove-destination \$(readlink \$f) \$f; done"

kubectl cp "<POD_NAME>:/workspace/kinc/output/" "output"

Open Cytoscape. (Applications -> Other -> Cytoscape)

Go to your desktop and open a file browsing window, navigate to the output folder:

cd ~/Desktop/classroom/myfiles/scidas-workshop/kinc/output

Finally, drag the file <DATASET>.coexpnet.txt from the file browser to Cytoscape!

The network should now be visualized!

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

Doing Genomics in the Cloud: From RNA to Result

0. Prerequisites

Install Helm

Access Praxis(optional)

Access Nautilus(optional)

Creating a Research Environment in Kubernetes

1. Access Kubernertes Cluster

2. Create Persistant Data Storage to Host Workflow Data

2a. Create a NFS Storage Class(optional)

2b. Create a Persistent Volume Claim(PVC)

3. Start a Data Transfer Pod

4. Grant Nextflow Permissions(optional)

GEMmaker

1. Index Arabidopsis Reference Genome(optional)

2. Pull RNA Data from NCBI

2a. Use a Built-In Workflow Utility

2b. Use a Single-Container Deployment to Pull Files Sequentially(optional)

2c. Use a Multi-Container StatefulSet to Pull Files in Parallel(optional)

2d. Use the Data Transfer Pod Utility to Pull Files Sequentially(optional)

4. Configure/Deploy GEMmaker

5. View Output

KINC

1. Configure KINC

1a. Load the Input Dataset(optional)

2. Deploy KINC

3. Retrieve and Visualize Gene Co-expression Network

About

Releases

Packages

Languages

Name		Name	Last commit message	Last commit date
Latest commit History 31 Commits
gemmaker		gemmaker
kinc		kinc
.gitignore		.gitignore
README.md		README.md
pvc.yaml		pvc.yaml

SciDAS/scidas-workshop

Folders and files

Latest commit

History

Repository files navigation

Doing Genomics in the Cloud: From RNA to Result

0. Prerequisites

Install Helm

Access Praxis(optional)

Access Nautilus(optional)

Creating a Research Environment in Kubernetes

1. Access Kubernertes Cluster

2. Create Persistant Data Storage to Host Workflow Data

2a. Create a NFS Storage Class(optional)

2b. Create a Persistent Volume Claim(PVC)

3. Start a Data Transfer Pod

4. Grant Nextflow Permissions(optional)

GEMmaker

1. Index Arabidopsis Reference Genome(optional)

2. Pull RNA Data from NCBI

2a. Use a Built-In Workflow Utility

2b. Use a Single-Container Deployment to Pull Files Sequentially(optional)

2c. Use a Multi-Container StatefulSet to Pull Files in Parallel(optional)

2d. Use the Data Transfer Pod Utility to Pull Files Sequentially(optional)

4. Configure/Deploy GEMmaker

5. View Output

KINC

1. Configure KINC

1a. Load the Input Dataset(optional)

2. Deploy KINC

3. Retrieve and Visualize Gene Co-expression Network

About

Resources

Stars

Watchers

Forks

Releases

Packages 0

Languages

Packages