Nextflow NeoEpitope predictions Pipeline

Pipeline takes fastq files from Tumor and Normal samples (WES or WGS) and optionally RNAseq from Tumor to predict neoepitopes.

The pipeline uses the following tools:

• MuTect2
• MuTect1
• VarScan2
• Strelka2/Manta
• Sequenza
• ASCAT
• CNVkit
• OptiType
• HLA-HD
• pVACseq (NetMHCpan NetMHCpanEL MHCflurry MHCflurryEL NetMHCIIpan NetMHCIIpanEL DeepImmuno)
• NeoFuse
• mixMHC2pred
• mixcr
• blastp

It outputs a vcf file with the annotated and filtered SNPs and Indels, which where called with each of the callers and a high confidence vcf file (hc) in which only variants that were called by a minimum of 2 of the callers are listed. All vcf files are annotated with VEP. In addition the germline variants are called using HaploTypeCaller and a phased vcf for pVACseq is generated as well. Copy number variations are analyzed using CNVkit, ASCAT, and sequenza. Tumor purity is estimated by ASCAT and Sequenza and is used to derive the clonality measure for the predicted neoantigens. Tumor mutational burden (TMB) is calculated for all variants over the entire read covered genome and for coding variants on read covered exons. HLA class I and class II alleles are predicted with OptiType and HLA-HD. Class I and Class II neoepitopes are predicted with pVACseq using netMHCpan, netMHCIIpan and mhcflurry. In addition mixMHC2pred is used as complement Class II neoepitope predictor. Fusion neoantigens are predicted with NeoFuse. All predicted neoeptiopes are searched via BLAST against protein databases (UniProt, RefSeq) to ensure that they are not present in a "non-aberrant" protein. CSiN immunogenicity score is reported for Class I, Class II and combined neoepitopes. A GBM model [1] is used to predict immunogenicity scores for MHC class I single nucleotide variant (SNV) neoantigens 8-11 amino acid residues in length. Finally mixcr is run to predict the TCR and BCR repertoire.

[1] https://github.com/vincentlaboratories/neoag/.

Hardware recommendations

nextNEOpi is designed to run on high memory multi-core servers (recommended > 16 cores, min. 64GB RAM, > 5 TB of free disk space). For analysis of larger cohorts we strongly recommend running nextNEOpi on a HPC cluster with multiple compute nodes that meet these specifications. However, by tuning the memory and CPU parameters in params.config and process.config it should also be possible to run nextNEOpi on systems with lower CPU and memory resources.

1. Installation

1.1 Nextflow

The command below may be used to install Nextflow. Please see also the installation instructions at: https://www.nextflow.io/index.html#GetStarted

curl -s https://get.nextflow.io | bash

1.2 Analysis tools and software packages

The pipeline will install almost all required tools via Singularity images or conda environments. If preferred one can also use local installations of all tools (not recommended, please see Manual installation at the end of this document)

The software that needs to be present on the system is Java (minimum version 8, if running conda java version 17 or higher is needed), Nextflow (see above), Singularity, Conda (optional).

If you intend to run the pipeline with the conda profile instead of singularity, we recommend to install mamba (https://github.com/mamba-org/mamba) to speed up the creation of conda environments. If you can not install mamba please set conda.useMamba = false for the conda profile in conf/profiles.config

Optional but recommended: Due to license restrictions you may also need to download and install HLA-HD by your own, and set the installation path in conf/params.config. If HLA-HD is not available Class II neoepitopes will NOT be predicted

1.3 References

The pipeline requires different reference files, indexes and databases:

please see conf/resources.config

For each nextNEOpi version we prepared a bundle with all needed references, indexes and databases which can be obtained from:

https://apps-01.i-med.ac.at/resources/nextneopi/

the bundle is named to match the release version nextNEOpi_<version>_resources.tar.gz

e.g.:

https://apps-01.i-med.ac.at/resources/nextneopi/nextNEOpi_1.4_resources.tar.gz

download and extract the contents of the archive into the directory you specified for resourcesBaseDir in the conf/params.config file.

The structure should look as shown blow:

├── {resourcesBaseDir}
    ├── databases
    ├── ExomeCaptureKits
    └── references

Notes

1. You may also provide your own versions of these files. To do so, please change the conf/resources.config accordingly.
2. Due to license restriction, we do not provide a copy of the optional COSMIC database. If you also want to include COSMIC data, you may get a copy at https://cancer.sanger.ac.uk/cosmic
3. We provide the region and bait files for two different Exome capturing kits from Agilent:
   • SureSelect Human All Exon V6 exome
   • SureSelect Human All Exon V7 exome
   • Twist Human comprehensive exome

You may add your own region and bait files by defining an entry in conf/resources.config

Refs:

• https://gatk.broadinstitute.org/hc/en-us/articles/360036212652-Resource-Bundle
• https://console.cloud.google.com/storage/browser/genomics-public-data/resources/broad/hg38/v0/
• ftp://[email protected]/bundle/
• https://gdc.cancer.gov/about-data/gdc-data-processing/gdc-reference-files
• https://www.gencodegenes.org/human/

1.4 Testdata

If you want to test the pipeline using a working minimal test dataset you may download one from

https://apps-01.i-med.ac.at/resources/nextneopi/nextNEOpi_testdata.tar.gz

Please note that due to the limited read coverage CNVkit will not run successfully using this test dataset. Please run the pipeline using the parameter --CNVkit false when testing with this dataset.

2. Usage

Before running the pipeline, the config files in the conf/ directory may need to be edited. In the params.config parameters default settings are defined. The process.config is a template for the configuration of the single processes, you may check the number of CPUs assigned for each process and adjust according to your systems capabilities.

Most pipeline parameters can be edited in the params.config file or changed on run time with command line options by using --NameOfTheParameter given in the params.config. References, databases should be edited in the resources.config file.

Note: nextNEOpi is currently written in nextflow DSL 1, which is only supported up to nextflow version 22.10.8, this means you need to pin the nextflow version by setting the environment variable NXF_VER=22.10.8, in case you have installed a newer nextflow version.

NXF_VER=22.10.8 nextflow run nextNEOpi.nf --batchFile <batchFile_FASTQ.csv | batchFile_BAM.csv> -profile singularity|conda,[cluster] [-resume] -config conf/params.config

Profiles: conda or singularity

We highly recommend to use either the singularity or conda profile. You can specify one of the two profiles using the option -profile singularity or -profile conda. This way you do not have to care about installing all the required software including all its dependencies.

Profiles: cluster

We strongly recommend to run the pipeline on a HPC cluster. You can enable runs in cluster mode by using a profile named e.g. cluster and the option -profile singularity,cluster or -profile conda,cluster

For an example Slurm cluster profile, please see profiles in conf/profiles.config. You may uncomment and adjust the cluster profile to your scheduling system.

Sequencing data input:

Besides raw reads in FASTQ fromated files, input data may also be provided in BAM format.

Note: Per default DNA sequencing data will be treated as exome sequencing WES data. If whole genome sequencing is used, please specify --WES false on the command line or set WES = false in params.config.

RNA reads from tag seq library i.e. 3-prime end sequencing protocol

--RNA_tag_seq turns off the "--trna-vaf" and "--trna-cov" filter from pVACseq epitope filtering. It also turns of HLA typing from RNAseq data. 3-prime end sequencing does not cover the entire transcript.

Mandatory arguments:

-config nextNEOpi config file, e.g. -config conf/params.config

--accept_license accept the license terms

--MIXCR_lic /path/to/mi.license OR --TCR false : a MiXCR license may be requested at https://licensing.milaboratories.com

--batchFile

Make sure that your batchFile CSV includes the column names as shown in the examples below as header line. See also example_batchFile_FASTQ.csv or example_batchFile_BAM.csv

FASTQ raw reads

• e.g.: CSV-file with Tumor/Normal WES/WGS, and optional RNAseq reads

sampleName	reads1	reads2	sampleType	HLAfile	sex
sample1	reads_s1_t_1.fastq.gz	reads_s1_t_2.fastq.gz	tumor_DNA		female
sample1	reads_s1_n_1.fastq.gz	reads_s1_n_2.fastq.gz	normal_DNA		female
sample1	reads_s1_r_1.fastq.gz	reads_s1_r_2.fastq.gz	tumor_RNA		female
sample2	reads_s2_t_1.fastq.gz	reads_s2_t2.fastq.gz	tumor_DNA	/data/sample2_hla.txt	male
sample2	reads_s2_n_1.fastq.gz	reads_s2_n_2.fastq.gz	normal_DNA		male
sample2	reads_s2_r_1.fastq.gz		tumor_RNA		male
sample3	reads_s3_t_2.fastq.gz	reads_s3_t_2.fastq.gz	tumor_DNA		female
sample3	reads_s3_n_1.fastq.gz	reads_s3_n_2.fastq.gz	normal_DNA		female
sample4	reads_s4_t_1_1.fastq.gz	reads_s4_t_1_2.fastq.gz	tumor_DNA		NA
sample4	reads_s4_t_2_1.fastq.gz	reads_s4_t_2_2.fastq.gz	tumor_DNA		NA
sample4	reads_s4_n_1.fastq.gz	reads_s4_n_2.fastq.gz	normal_DNA		NA
sample4	reads_s4_r_1.fastq.gz	reads_s4_r_2.fastq.gz	tumor_RNA		NA
...

• CSV fields of the batchFile
• reads1: forward reads (can be fastq or gzipped fastq)
• reads2: reverse reads (if paired end sequencing was used, empty for single end)
• sampleType: one of tumor_DNA, normal_DNA. tumor_RNA
• HLAfile: optional file with HLA types (default: empty, example see nextNEOpi_testdata.tar.gz)
• sex: gender of the sample if known (female, male, xx, xy) or NA if unknown

A sample may have multiple read files for a single sampleType, nextNEOpi will merge them accordingly. As shown in the above example sample4 has 2 fastq files for the tumor_DNA, in this cases reads_1_1.fastq.gz will be merged with reads_2_1.fastq.gz. The same applies to reads2.

BAM files

Note: If BAM files are used it is very much recommended that they also include also the unmapped and multimapping reads. These reads can be helpful for HLA-typing.

• e.g.: CSV-file with Tumor/Normal WES/WGS, and optional RNAseq data:

sampleName	bam	sampleType	HLAfile	sex
sample1	sample1_t.bam	tumor_DNA		female
sample1	sample1_n.bam	normal_DNA		female
sample1	sample1_r.bam	tumor_RNA		female
sample2	sample2_t.bam	tumor_DNA		male
sample2	sample2_n.bam	normal_DNA		male

Notes

• You can not use samples that have mixed single-end and paired-end DNA reads in tumor and normal. Both, tumor and normal DNA library types need to be either SE or PE for a given sample

• in the HLAfile coulumn a user suppiled HLA types file may be specified for a given sample

• the sex column can be "XX", "female" or "Female", "XY", "male" or "Male". If not specified or "NA" the gender is inferred from the data.

• Please make sure your /tmp (or the directory set as $TMPDIR) has a lot of free space (> 50GB, depending on input data size). You may change the tmp dir used by nextNEOpi by setting the parameter tmpDir in params.config or on the command line --tmpDir to a directory of your choice.

Example run command with batchfile:

nextflow run nextNEOpi.nf \
    --batchFile batchfile.csv \
    -config conf/params.config \
    --outputDir /data/results/nextNEOpi/myResults \
    --trim_adapters true \
    --trim_adapters_RNAseq true \
    --use_NetChop false \
    --tmpDir /my/huge/tmpDir \
    -profile singularity,cluster \
    -resume

Optional argument:

--trim_adapters If true adapter sequences are automatically determined and will be trimmed from reads. If --adapterSeq (string of adapter sequence) or --adapterSeqFile (fasta file with adapter sequences) is provided then adapters will be used as specified (no automatic detection). Default: false

--trim_adapters_RNAseq If true adapter sequences are automatically determined and will be trimmed from RNAseq reads. If --adapterSeqRNAseq (string of adapter sequence) or --adapterSeqFileRNAseq (fasta file with adapter sequences) is provided then adapters will be used as specified (no automatic detection). Default: false

--adapterSeq String of adapter sequence (see --trim_adapters) --adapterSeqFile Fasta file with adapter sequence(s) (see --trim_adapters)

--adapterSeqRNAseq String of adapter sequence (see --trim_adapters_RNAseq) --adapterSeqFileRNAseq Fasta file with adapter sequence(s) (see --trim_adapers_RNAseq)

--mutect2ponFile Panel of Normals file for Mutect2 (https://gatk.broadinstitute.org/hc/en-us/articles/360035890631-Panel-of-Normals-PON-) Default: false

--priorityCaller Set the variant caller used as base for the hc variants. Only variants that are confirmed by any of the two confirming callers (e..g. mutect1, varscan) will be retained. M2 = mutect2, M1 = mutect1, VS = varscan, ST = strelka Default: M2

--minAD Minimum allelic depth (reads covering a variant) Default: 5

--use_NetChop Use NetChop to generate peptides Default: false

--use_NetMHCstab Use NetMHCstab to predict the stability of peptide binding to MHC molecules Default: true

--TCR Run mixcr for TCR prediction Default: true

--CNVkit Run CNVkit for detecting CNAs. Default: true

--HLAHD_DIR Specify the path to your HLA-HD installation. Needed if Class II neoantigens should be predicted.

--HLA_force_RNA Use only RNAseq for HLA typing. Default: false

--HLA_force_DNA Use only WES/WGS for HLA typing. Default: false

--run_HLAHD_RNA Run HLA-HD also on RNAseq. Highly accurate but can be very slow on larger fastq files. Default: false

--disable_OptiType Disable OptiType for HLA typing. If set, HLA-HD or a user supplied custom HLA file must be available (see --HLAHD_DIR and/or HLAfile field in the batchFile)

--pVACseq_filter_set Can be one of [standard, relaxed, custom]. The standard filter set is using the pVACseq default filters. The relaxed filter set is filtering only for ic50 < 500 & rank < 2 & expn-val > 2. With filter set custom users can define a custom set of filters by providing the desired filters (space separated) using the --pVACseq_custom_filters option. E.g. --pVACseq_filter_set custom --pVACseq_custom_filters "--binding-threshold 250 --percentile-threshold 1". For filter options please see also the pVACseq manual. Default: standard

--pVACseq_custom_filters See --pVACseq_filter_set

Further options: There are many more options that can be set in the params.conf file or specified on the commandline (see conf/params.config)

3. Output

The Pipeline stores its ouput in the following structure:

RESULTS
├── analyses
│   ├── Subject_01
│   │   ├── 01_preprocessing
│   │   ├── 02_alignments
│   │   ├── 03_baserecalibration
│   │   ├── 03_realignment
│   │   ├── 04_expression
│   │   ├── 04_variations
│   │   ├── 05_vep
│   │   ├── 06_proteinseq
│   │   ├── 07_MutationalBurden
│   │   ├── 08_CNVs
│   │   ├── 09_CCF
│   │   ├── 10_HLA_typing
│   │   ├── 11_Fusions
│   │   ├── 12_pVACseq
│   │   ├── 13_mixMHC2pred
│   │   ├── 14_CSiN
│   │   ├── 14_IGS
│   │   ├── 15_BCR_TCR
│   │   └── QC
│   ├── Subject_02
│   │   ├── [...]
│   ├── [...]
│   │   ├── [...]
│   ├── Subject_n
│   │   ├── [...]
├── Documentation
├── neoantigens
│   ├── Subject_ID
│   │   ├── Class_I
│   │   ├── Class_II
│   │   └── Final_HLAcalls
│   ├── Subject_02
│   │   ├── [...]
│   ├── [...]
│   │   ├── [...]
│   ├── Subject_n
│   │   ├── [...]
├── pipeline_info
│   └── icbi
└── supplemental
    ├── 00_prepare_Intervals
    └── 01_prepare_CNVkit

4. [Manual installaton: Not recommended]:

If you prefer local installation of the analysis tools please install the following software:

• FASTQC (Version >= 0.11.8)
• FASTP (Version >= v0.20.1)
• JAVA7 (Version 1.7)
• JAVA8 (Version => 1.17)
• BWA (Version >= 0.7.17)
• SAMTOOLS (Version >= 1.9)
• GATK3 (Version 3.8-0)
• GATK4 (Version >= 4.5.0.0)
• VARSCAN (Version 2.4.6)
• MUTECT1 (Version 1.1.7) ---- optional
• BAMREADCOUNT (Version 0.8.0)
• VEP (Version v111)
• BGZIP
• TABIX
• BCFTOOLS
• MANTA
• STRELKA
• SAMBAMBA
• OPTITYPE
• PYTHON
• PERL
• CONDA
• YARA
• HLA-HD
• ALLELECOUNT
• RSCRIPT (R > 3.6.2)
• SEQUENZA (3.0)
• CNVkit

all these tools need be available via the $PATH environment variable. However, you still need Java, Nextflow, Conda and Singularity installed on your system.