ImmuneMirror: A Machine Learning-based Integrative Pipeline and Web Server for Neoantigen Prediction

Overview

We developed ImmuneMirror, a multi-omics data analysis bioinformatics pipeline to access the key genomic and transcriptomic features associated with the response of cancer immunotherapy. The pipeline was built as a docker container that can be run in any docker supported operating system such as Linux, Mac and Windows. We incorporated a machine-learning model in the pipeline, to evaluate each mutated peptide and aggregates significant biological features for neoantigen prediction and prioritization. The analysis pipeline incorporates the benchmark tools for identifying the germline and somatic mutations, evaluation of microsatellite instability (MSI), HLA typing, and neoantigen prediction and for HLA Class I and II based on the whole-exome sequencing (WES) and RNA-Seq data. The pipeline required FASTQ input of matched normal-tumor WES samples and tumor bulk RNA-seq sample. In addition, the pipeline produces a two-page long graphical analysis report for the user sample(s).

Abstract

We developed ImmuneMirror as stand-alone open-source pipeline (https://github.com/weidai2/ImmuneMirror/), it's required FASTQ input of matched normal-tumor WES samples and tumor bulk RNA-seq sample (RNA-seq is optional). If you have VCF file (somatic mutations identified by Mutect2), you can use the ImmuneMirror web server (http://immunemirror.hku.hk/App/).

System Requirements

* Docker version: Platform of your choice.
- Bash version: Linux, we run this pipeline under Ubuntu 20.04 LTS.
- Hardware:
    Minimum: 16-core processor, 64 GB RAM
    Optimal: 32-core processor, 128 GB RAM
- Disk space:
    Docker image: 79.6 GB 
    Reference files: 483 GB
    Results: around 41 GB, for one patient sample

How to use:

1. Download and install Docker

2. Download the Docker image from this link: http://immunemirror.hku.hk

   Now, you need to load the docker image to your docker environment (local machine) by executing the following command:

   docker load < immunemirror-1.0.tar
   

3. Download and unzip the pipeline's repository from Github:

Way 1:

Copy im_install.sh to your local computer and set you working directory's path inside the im_install.sh script by editing the following line:

working_directory=/Porvide/PATH/TO/YOUR_WORKING_DIRECTORY 

Now, run the script below (im_install.sh):

chmod +x im_install.sh\ 
&& ./im_install.sh

Way 2:

You can directly clone the master repository from the GitHub, unzip it and rename as ImmuneMirror

Now your working environment is ready!

4. Download the reference files and example samples from the link below, and unzip them: http://immunemirror.hku.hk

5. Test the pipeline using example samples. It will take approximately 24 hours, Depending on the available CPU and RAM.

   You need to map the local directories to the Docker container as follows:

   WES_directory=/PATH/TO/YOUR/WES_Seq_DIRECTORY
   RNASeq_directory=/PATH/TO/YOUR/RNASeq_DIRECTOR
   Reference_files_directory=/PATH/TO/Reference_file_directory
   

Now, run the commands below to process the example samples:

disease_type=CRC \
thread=13 \
   sudo docker run \
    -v {your_working_directory}/ImmuneMirror/:/var/pipeline/ \
    -v {WES_directory}:/var/pipeline/WES \
    -v {RNASeq_directory}:/var/pipeline/RNASeq \
    -v {Reference_files_directory}/:/var/pipeline/Ref/ \
     immunemirror:1.0 ./ImmuneMirror.sh $disease_type $thread 

6. Now run can the pipeline using "real-life" samples.

   Firstly, you need to edit the {working_directory}/ImmuneMirror/sample.list file by replacing with your own sample list. Inside the 'sample.list' file, "YES" indicates the patient has both WES and RANSeq files, and "NO" indicates the patient has only WES sequecing file. 'N' and 'T' represents Normal and Tumor sample, respectively.

   Available disease types: ESCC (Esophageal squamous cell carcinoma), CRC (Colorectal cancer), and HCC (Hepatocellular carcinoma), and No for other cancer/disease types. Now, run the following commnads to process you samples:

disease_type={your_sample_disease_type} \
thread=13 \
    sudo docker run \
    -v {your_working_directory}/ImmuneMirror/:/var/pipeline/ \
    -v {WES_directory}:/var/pipeline/WES \
    -v {RNASeq_directory}:/var/pipeline/RNASeq \
    -v {Reference_files_directory}/:/var/pipeline/Ref/ \
     immunemirror:1.0 ./ImmuneMirror.sh $disease_type $thread

Bug reports

Please send comments and bug reports to: sarwar20@hku.hk

Citation

Chuwdhury GS, Guo Y, Chiang CL, Lam KO, Kam NW, Liu Z, Dai W. ImmuneMirror: A machine learning-based integrative pipeline and web server for neoantigen prediction. Brief Bioinform. 2024 Jan 22;25(2):bbae024. doi: 10.1093/bib/bbae024. PMID: 38343325; PMCID: PMC10859690.

License

The program is distributed under the GPL-3.0 license.