Neoantigens derived from somatic DNA alterations are ideal cancer-specific targets. However, not all somatic DNA mutations can result in immunogenicity in cancer cells, and efficient tools for predicting the immunogenicity of neoepitope are still urgently needed. Here we present the Seq2Neo pipeline, which provides a one-stop solution for neoepitope features prediction from raw sequencing data, and neoantigens derived from different types of genome DNA alterations, including point mutations, insertion deletions, and gene fusions are supported. Importantly a convolutional neural networks (CNN) based model has been trained to predict the immunogenicity of neoepitope. And this model shows improved performance compared with currently available tools in immunogenicity prediction in independent datasets.
Seq2Neo runs on a Linux operation system like the CentOS system (recommended), and it is open-source software under an academic free license (AFL) v3.0.
We strongly recommend using the conda command line for installation as this will solve dependencies automatically. The web of the package is https://anaconda.org/liuxslab/seq2neo.
-
Firstly, you need to install the Anaconda or Miniconda (recommended), and set channels in the
~/.condarcfile like this:channels: - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/bioconda - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/menpo - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/r - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/msys2 - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/simpleitkYou can replace Tsinghua mirrors with other convenient mirrors.
-
Secondly, you should execute the following commands to create a new environment named Seq2Neo or other on your Linux system, and then activate it:
conda create -n Seq2Neo conda activate Seq2Neo
-
Thirdly, you can install the package through the following conda command:
conda install -c liuxslab seq2neo
-
Finally, please installation of following packages manually due to the reasons of permission or others:
- Annovar == latest ANNOVAR website (openbioinformatics.org)
- HLAHD == 1.4.0 HLA-HD (kyoto-u.ac.jp)
- netCTLpan == 1.1.b NetCTLpan - 1.1 - Services - DTU Health Tech
- netMHCpan == 4.1.b NetMHCpan - 4.1 - Services - DTU Health Tech
- STAR-Fusion == 1.10.1 STAR-Fusion/STAR-Fusion: STAR-Fusion codebase (github.com)
Following corresponding official instructions to install those packages on your system.
We also provide docker image (liuxslab/seq2neo - Docker Image | Docker Hub) that contains all packages dependencies. You need to install docker in advance on your system. Then the command docker pull liuxslab/seq2neo:latest will pull the seq2neo image into your local machine. You can put resource files required by BWA, Mutect2, and others in one folder resource_files, which has several classified folders like bqsr_resource, mutect2_resource, starfusion_resource, ref_genome (a reference to the section of "The module of whole"), then execute the following commands to start a docker container and activate Seq2Neo conda environment including seq2neo and its dependencies:
docker run -it -v /path/to/resource_files:/home/resource_files liuxslab/seq2neo:latest /bin/bash
cd /home/seq2neo
cd biosoft/hlahd.1.4.0/ && sh install.sh && cd ../../ # installation of HLAHD 1.4.0
conda activate Seq2Neo
In the Seq2Neo environment, you can run seq2neo commands, please refer to the following section of "The module of whole".
You can install the stable release of Seq2Neo with:
pip install Seq2Neo
However, you should install all of the dependencies manually. It includes the following softwares and packages that should be installed in advance:
- bamtools=2.5.1
- bwa=0.7.17
- fastp=0.23.2
- perl=5.26.2=h470a237_0
- samtools=1.15.1
- star=2.7.8a
- tpmcalculator=0.0.4
- vcftools=0.1.16
- bowtie2 == 2.3.5
- gatk == 4.2.5
Then, you should also install the packages mentioned in the Conda section.
Seq2Neo consists of 3 modules, which are whole, download, and immuno. The module of whole is responsible for running the entire process, the download module can download a specified version of a reference genome from the Ensembl database and index, and the last module of immuno supports the prediction of immunogenicity score of specified peptides and MHCs:
usage: seq2neo [-h] {whole,immuno,download} ...
A pipeline from mutation to neoantigen prediction
positional arguments:
{whole,immuno,download}
whole Run whole pipeline (Seq2Neo) with fastq/bam/sam file
immuno Run immunogenicity prediction with peptides
download downloading reference genome
optional arguments:
-h, --help show this help message and exit
Thanks for using Seq2Neo
The module of whole:
usage: seq2neo whole [-h] [--data-type {sort_bam,sam,fastq}] --ref
path_to_reference
[--normal-dna normal_dna_1.fq normal_dna_2.fq]
[--tumor-dna tumor_dna_1.fq tumor_dna_2.fq] --tumor-rna
tumor_rna_1.fq tumor_rna_2.fq [--normal-sam normal.sam]
[--tumor-sam tumor.sam]
[--normal-sorted-bam normal_sorted.bam]
[--tumor-sorted-bam tumor_sorted.bam] --normal-name
normal_name --tumor-name tumor_name --known-site-dir
known_site_dir --mutect2-dataset-dir mutect2_dataset_dir
--annovar-db-dir annovar_db_dir --genome-lib-dir
genome_lib_dir --agfusion-db agfusion_db [--out out_dir]
[--len [LEN [LEN ...]]] [--threadN thread_num]
[--mdna min_length] [--mrna min_length]
[--java-options java_options] --hlahd-dir hlahd_dir
[--hla [HLA [HLA ...]]]
Run whole pipeline (Seq2Neo) with fastq/bam file
optional arguments:
-h, --help show this help message and exit
--data-type {sort_bam,sam,fastq}
Select your input file format(fastq/sam/sort_bam),
default:fastq (default: fastq)
--ref path_to_reference
Path to reference genomic data (default: None)
--normal-dna normal_dna_1.fq normal_dna_2.fq
Normal sample files (default: None)
--tumor-dna tumor_dna_1.fq tumor_dna_2.fq
Tumor dna sample files (default: None)
--tumor-rna tumor_rna_1.fq tumor_rna_2.fq
Tumor rna sample files (default: None)
--normal-sam normal.sam
Normal dna sam files (default: None)
--tumor-sam tumor.sam
Tumor dna sam files (default: None)
--normal-sorted-bam normal_sorted.bam
Normal dna sorted bam files (default: None)
--tumor-sorted-bam tumor_sorted.bam
Tumor dna sorted bam files (default: None)
--normal-name normal_name
if the file is XXX_1.fq, the normal name should be XXX
(default: None)
--tumor-name tumor_name
if the file is XXX_1.fq, the tumor name should be XXX
(default: None)
--known-site-dir known_site_dir
directory to BQSR known sites (default: None)
--mutect2-dataset-dir mutect2_dataset_dir
directory to mutect2 needed dataset file (default:
None)
--annovar-db-dir annovar_db_dir
directory to annovar database (default: None)
--genome-lib-dir genome_lib_dir
directory containing genome lib (see http://STAR-
Fusion.github.io) (default: None)
--agfusion-db agfusion_db
Path to the AGFusion database (default: None)
--out out_dir Output directory to save prediction results, default
is current directory (default: .)
--len [LEN [LEN ...]]
length of peptides, default is 8 9 10 11 (default: (8,
9, 10, 11))
--threadN thread_num the number of thread used in Seq2Neo, default is 4
(default: 4)
--mdna min_length length of reads(normal dna) (default: 100)
--mrna min_length length of reads(rna) (default: 100)
--java-options java_options
set config for java (default: -Xmx8G)
--hlahd-dir hlahd_dir
the path to hlahd software (default: None)
--hla [HLA [HLA ...]]
if you use bam and sorted sam as input, please provide
hlas, like--hla HLA-A01:01 HLA-A03:03 (default:
HLA-A01:01)
-
You need to download the necessary reference files before running Seq2Neo:
-
Downloading three BQSR known sites files used to recalibrate base quality score, those files should be put in a directory like bqsr_resource, and index files are needed to accelerate the speed of Seq2Neo. The commands are following:
mkdir bqsr_resource && cd bqsr_resource prefix=ftp://gsapubftp-anonymous@ftp.broadinstitute.org/bundle/hg38/ wget ${prefix}dbsnp_146.hg38.vcf.gz wget ${prefix}dbsnp_146.hg38.vcf.gz.tbi wget ${prefix}1000G_phase1.snps.high_confidence.hg38.vcf.gz wget ${prefix}1000G_phase1.snps.high_confidence.hg38.vcf.gz.tbi wget ${prefix}Mills_and_1000G_gold_standard.indels.hg38.vcf.gz wget ${prefix}Mills_and_1000G_gold_standard.indels.hg38.vcf.gz.tbi -
Downloading hg38 datasets of annovar via the following commands:
cd /path/to/annovar perl annotate_variation.pl --downdb --webfrom annovar --buildver hg38 refGene humandb/ -
Downloading the necessary reference files used to call Mutect2, those files should be put in a directory like mutect2_resource, and index files are needed to accelerate the speed of Seq2Neo. The commands are following:
mkdir mutect2_resource && cd mutect2_resource prefix=ftp://gsapubftp-anonymous@ftp.broadinstitute.org/bundle/Mutect2/ wget ${prefix}af-only-gnomad.hg38.vcf.gz wget ${prefix}af-only-gnomad.hg38.vcf.gz.tbi wget ${prefix}GetPileupSummaries/small_exac_common_3.hg38.vcf.gz wget ${prefix}GetPileupSummaries/small_exac_common_3.hg38.vcf.gz.tbi -
Downloading the AGFusion database and pyensembl reference genome, we select the max release of 95 to download:
pyensembl install --species homo_sapiens --release 95 agfusion download -g hg38 --release 95 -
Downloading the genome library of STAR-Fusion (1.10.1) to call gene fusions via the following commands:
ref=GRCh38_gencode_v37_CTAT_lib_Mar012021.plug-n-play.tar.gz wget https://data.broadinstitute.org/Trinity/CTAT_RESOURCE_LIB/__genome_libs_StarFv1.10/${ref} tar -zxvf ${ref}The size of the compressed genome library is about 31 G, Chinese researchers can download it at a higher speed by using some useful tools like Thunder Official Website.
-
Downloading the reference genome and indexing via the following command lines:
wget https://github.com/broadinstitute/gatk/raw/master/src/test/resources/large/Homo_sapiens_assembly38.fasta.gz tar -zxvf Homo_sapiens_assembly38.fasta.gz bwa index -a bwtsw Homo_sapiens_assembly38.fasta ## build index for Homo_sapiens_assembly38.fasta samtools faidx Homo_sapiens_assembly38.fasta gatk CreateSequenceDictionary -R Homo_sapiens_assembly38.fasta -O Homo_sapiens_assembly38.dict
-
-
Suppose you have downloaded three files, they are tumor RNA-seq and WES data, normal WES data. Specifically, SRR2603346 is for tumor RNA-seq, SRR2601737 is for tumor WES, and SRR2601758 is for normal WES. Then you can run Seq2Neo via the following command line to obtain potential neoantigens (running on a machine with more than 50G memory and 512G hard disk space):
seq2neo whole --ref ref_genome/Homo_sapiens_assembly38.fasta --normal-dna SRR2601758_1.fastq SRR2601758_2.fastq --tumor-dna SRR2601737_1.fastq SRR2601737_2.fastq --tumor-rna SRR2603346_1.fastq SRR2603346_2.fastq --normal-name SRR2601758 --tumor-name SRR2601737 --known-site-dir bqsr_resource/ --mutect2-dataset-dir mutect2_resource/ --annovar-db-dir /path/to/annovar/humandb/ --genome-lib-dir /path/to/GRCh38_gencode_v37_CTAT_lib_Mar012021.plug-n-play/ctat_genome_lib_build_dir/ --agfusion-db agfusion.homo_sapiens.95.db --out out/ --len 8 9 10 11 --threadN 20 --java-options '"-Xmx40G"' --hlahd-dir /path/to/hlahd -
The final result of the module whole is in the folder of final_result, including final_results_neo.txt and filtered_neo.txt. The final_results_neo.txt includes all peptides from the detected mutation sites. After applying the criteria of TAP>0, TPM>0, immunogenicity>0.5 and IC50<=500, filtered_neo.txt is acquired (ranking by IC50).
The module of download:
usage: seq2neo download [-h] [--species SPECIES] [--build BUILD]
[--release RELEASE] [--dir [DIR]]
Run download module
optional arguments:
-h, --help show this help message and exit
--species SPECIES which species to download (default: homo_sapiens)
--build BUILD which build to download (default: GRCh38)
--release RELEASE which release to download (default: 105)
--dir [DIR] where to store (default: .)
This module will help users download and index reference genomes from the Ensembl database. The usage of the module is:
seq2neo download --species homo_sapiens --build GRCh38 --release 105 --dir .
The module of immuno:
usage: seq2neo immuno [-h] [--mode MODE] [--epitope EPITOPE] [--hla HLA]
[--inputfile INPUTFILE] [--outdir OUTDIR]
Seq2Neo-CNN command line(one part of Seq2Neo)
optional arguments:
-h, --help show this help message and exit
--mode MODE single mode or multiple mode (default: single)
--epitope EPITOPE if single mode, specifying your epitope (default:
SVQIISCQY)
--hla HLA if single mode, specifying your HLA allele (default:
HLA-A30:02)
--inputfile INPUTFILE
if multiple mode, specifying the path to your input
file (default: None)
--outdir OUTDIR if multiple mode, specifying the path to your output
folder (default: None)
The module allows users to predict the immunogenicity scores of provided peptides and HLAs.
For example, if you want to query a single peptide SVQIISCQY along with HLA-A30:02. You need to type:
seq2neo immuno --mode single --epitope SVQIISCQY --hla HLA-A30:02
If you want to query multiple epitopes, you just need to prepare a csv format file like this:
Pep,HLA
ADTSEARPFW,HLA-B44:02
ADVLSPVLVK,HLA-A03:01
AELEEVSSY,HLA-B44:02
AELLAKQLY,HLA-B44:02
AEQQGACPGL,HLA-B44:02
AEVSVLYTV,HLA-B44:02
AEYQDMHSY,HLA-B44:02
AINRPTVLK,HLA-A03:01
Then you run:
seq2neo immuno --mode multiple --inputfile data/test_input.csv --outdir data/
You will get two files, immuno_input_file.csv and cnn_results.csv. The former includes the predictions of TAP and IC50 performed by netCTLpan and netMHCpan4.1b, respectively, and the latter is the final results including immunogenicity scores.