ImmunoPepper

ImmunoPepper is a software tool that takes a splicing graph (possibly derived from an RNA-Seq samples) as input and generates the set of all theoretically peptide sequences (or kmers) through direct translation of all walks along the graph. This peptide set can be personalized with germline and somatic variants and takes all exon-exon junctions present in the splicing graph (even ones not part of the reference annotation but present in the given RNA-Seq sample) into account. The comprehensive set of peptides can be used subsequently for further downstream analyses such as domain annotation or computational immunology.

Get Started

Installation

It is recommended to setup a separate virtual or conda environment. The basic ImmunoPepper package can be installed via pip:

pip install immunopepper

Alternatively, ImmunoPepper can also be installed from source using:

pip install -r requirements.txt -r requirements_dev.txt
make install

After installation, please consult the help screen for further usage options:

immunopepper -h

Prerequisites

ImmunoPepper takes a splicing graph as input. This splicing graph has to be generated using the SplAdder pipeline. Further information about SplAdder is available on its GitHub page or the Online documentation.

Basic workflow

The software has four basic working modes:

1. build: Core part of ImmunoPepper. Traverses the input splice graph and generates all possible peptides/kmers.
2. make_bg: Integrates multiple kmer files (produced via build) and generates one background kmer file.
3. diff: Takes as input the foreground kmer file and a background kmer file. The output is contrasting foreground and background, indicating all foreground kmers not present in the background.
4. filter: Apply different filter mechanisms to a given kmer file.

Mode build

The following parameters are mandatory:

• --samples: input sample names; can specify more than one sample. (Example: 'sample1 sample2')
• --output-dir: output directory
• --ann-path:annotation file, accepted file formats: .gtf, .gff and .gff3
• --splice-path: path of the input SplAdder splice graph
• --ref-path: reference genome file in FASTA format

The following parameters are optional:

• --mutation-mode: mutation mode; choose from {ref, germline, somatic and somatic_germline}, default mode ref.
• --kmer: length of the kmers for kmer ouput. Default value is 0, which will output full peptides instead of kmers. A recommended kmer length is 9.
• --disable-concat: Turns off the generation of kmers from combinations of more than 2 exons (kmers generated from combinations of short exons might be missed)
• --germline: germline mutation file path. Mandatory argument if the mutation mode is germline or somatic_and_germline.
• --somatic: somatic mutation file path. Mandatory argument if the mutation mode is somatic or somatic_and_germline.
• --use-mut-pickle: Summarize mutation information in a pickle file and re-use if existing (saves the time processing the original mutation files).
• --count-path: path to splice graph count file
• --compressed: compress the output files using gzip

Example command line (replace ref with germline to consider mutation information)

immunopepper build \
--output-dir tests/test1/current_output_pos \
--ann-path tests/test1/data/test1pos.gtf \
--ref-path tests/test1/data/test1pos.fa \
--splice-path tests/test1/data/posgraph/spladder/genes_graph_conf3.merge_graphs.pickle \
--somatic tests/test1/data/test1pos.maf \
--germline tests/test1/data/test1pos.vcf \
--samples test1pos test1neg \
--mutation-mode ref \
--kmer 4 \
--disable-concat \
--count-path  tests/test1/data/posgraph/spladder/genes_graph_conf3.merge_graphs.count.hdf5

Mode make_bg

The following parameters are mandatory:

• --kmer-files: The list of kmer files output by build mode, e.g., 'ref_back_kmer.txt somatic_back_kmer.txt'.
• --output-file-path: Output integrated background kmer file path.
• --output-dir: Directory to store the log file.

The following parameters are optional:

• --verbose: Specify the level of output. 0 means zero debug information, 2 means the most detailed information.
• --compressed: Compress the output files with gzip.

Example command line:

immunopepper make_bg \
--kmer-files-list tests/test1/current_output_pos/test1pos/ref_back_kmer.txt tests/test1/current_output_pos/test1pos/germline_back_kmer.txt \
--output-dir tests/test1/current_output_pos/ \
--output-file_path tests/test1/current_output_pos/test1pos/uniq_back_kmer.txt \
--verbose 2

Mode diff

The following parameters are mandatory:

• --junction-kmer-file: foreground junction file path generated by build mode, e.g., ref_junction_kmer.txt
• --bg-file-path: background kmer file path. Can be the output of make_bg mode or external file. One kmer per line.
• --output-file-path: output tsv file path.
• --output-dir: directory to store the log file.

The following parameters are optional:

• --verbose: Specify the verbosity level of output. 0 means zero debug information, 2 means the most detailed information.
• --compressed: compress the output files with gzip.

Example command line

immunopepper  diff \
--junction-kmer-file tests/test1/current_output_pos/test1pos/ref_junction_kmer.txt \
--bg-file-path tests/test1/current_output_pos/test1pos/uniq_back_kmer.txt \
--verbose 1 \
--output-file-path tests/test1/current_output_pos/test1pos/kmer_result.tsv \
--output-dir tests/test1/current_output_pos \
--remove-bg

Mode filter

The following parameters are mandatory:

• --junction-kmer-tsv-file: The original kmer tsv files. Generated by build mode or by diff mode. It should contain field cross-junction, seg-expr and junc_expr.

The following parameters are optional:

• --cross-junction: Only output the cross-junction kmers.
• --seg-expr: Only output kmers that have segment expression greater than threshold.
• --seg-expr-thresh: Segment expression threshold. Default 0.
• --junc-expr: Only output kmers that have junction expression greater than threshold.
• --junc-expr-thresh: Junction expression threshold. Default 0.
• --output-file-path: Mandatoray argument. Specify the output tsv file path.
• --output-dir: Mandatoray argument. Specify the directory to store the log file.
• --verbose: Specify the level of output. 0 means zero debug information, 2 means the most detailed information.
• --compressed: Compress the output files with gzip.

Example command line

immunopepper filter \
--junction-kmer-tsv_path tests/test1/current_output_pos/test1pos/kmer_result.tsv \
--output-dir tests/test1/current_output_pos/ \
--output-file-path tests/test1/current_output_pos/test1pos/kmer_result_filtered.tsv \
--cross-junction \
--junc-expr \
--verbose 2

post-processing guidlines

For further filtering, the user can use the predicted kmers as input for MHC-binding prediction or use MS databases for further confirmation.

MHC-Binding

One option for MHC binding prediction is NetMHC. Using the predicted kmers as input, NetMHC predicts a peptide-MHC class 1 binding score for each sequence using a neural network.

Mass spectrometry

Mass spectrometry data can provide further evidence for the presence of a predicted peptide. There exist several tools for searching a peptide sequence against a MS database, for instance OpenMS.

Output files

There are 5 files for the build mode. mut_mode refers to ref, somatic, germline and somatic_and_germline.

• [mut_mode]_back_peptides.fa: Peptides translated from annotation transcripts. Two lines for one output. The first line is the transcript ID and the second line is the result peptide.
• [mut_mode]_back_kmer.txt: kmers generated from [mut_mode]_back_peptides.fa. There are four columns: [ kmer, gene_name, seg_expr, is_crossjunction]. The first column is the result kmer, the second column is the transcript ID, the third column is the average segment expression and the final column is the flag indicating if the kmer is junction kmer. The final column is False for all rows in this file.
• [mut_mode]_peptides.fa: Peptides translated from traversing splicegraph. Two lines for one output. The first line is the output ID and the second line is the result peptide.
• [mut_mode]_junction_kmer.txt: kmers generated from [mut_mode]_peptides.fa. In addition to the same four columns in [mut_mode]_back_kmer.txt, there is one more column in this file. junction_expr, refers to the junction counts for those kmers that span over exon junction. For those with junction_expr > 0, the flag is_crossjunction is True.
• [mut_mode]_metadata.tsv.gz: Contain details for every junction pairs.

Detail explanation for columns in [mut_mode]_metadata.tsv.gz

• output_id: In the format of [gene_nama]:[first vertex]_[second vertex]:[somatic variant combination id]:[read frame]. Like GENE1:0_2:0:1. GENE1 is the gene name, 0_2 means this junction consists of vertex 0 and vertex 2. 0 means there is no somatic mutation or it is the first case of all somatic mutation combination cases. 2 means the read frame is 2.
• read_frame: int (0,1,2). The number of base left to the next junction pair.
• gene_name: str. The name of Gene.
• gene_chr: str. The Chromosome id where the gene is located.
• gene_strand: str ('+', '_'). The strand of gene.
• mutation_mode: str ('ref', 'somatic', 'germline', 'somatic_and_germline'). Mutation mode
• peptide_annotated: Boolean. Indicate if the junction peptide also appears in the background peptide.
• junction_peptided: Boolean. Indicate if the junction also appear in the input annotation file.
• has_stop_codon: Boolean. Indicate if there is stop codon in the junction pair.
• is_in_junction_list: Boolean. Indicate if the junction pair appear in the given junction whitelist.
• is_isolated: Boolean. Indicate if the output peptide is actually translated from a single exon instead of two.
• variant_comb: shows the somatic variantion combination used in this line of output. seperate by ';' eg. 5;25 means the somatic mutation of position 5 and 25 take effect in this output.
• variant_seg-expr: shows the corresponding expression of segments where the corresponding somatic mutation is in. eg. 257.0;123.2 means the segment where the somatic mutation in position 5 is in has counts 257.0
• modified_exons_coor: Shows exon coordination. Usually we have 4 number start_v1;stop_v1;start_v2;stop_v2. They have already absorb reading frame so you can use the coord directly to generate the same output peptide.
• original_exons_coord: Shows the original exon coordination.
• vertex_idx: shows the vertex id of the given junction. eg 5,6 means this junction pair consists of the fifth and sixth vertex.
• junction_expr: float. The expression of the junction.
• segment_expr: float. The weighted sum of segment expression. We split the junction into segments and compute the segment expression with the length-weighted-sum expression.

The .meta file is compressed by default in all time. The user can add --compressed option in the input argument to have other files compressed. It is recommended to output in compressed format because it can save a lot of storage.

The output file for make_bg mode is a text file. Each line is a unique kmer.

The output file for diff mode is a text file. There is a header line like [mut_mode]_junction_kmer but with one more column is_neo_flag to indicate if the kmer also exist in the background kmer file. We can also remove those kmers that exist in the background files with the option --remove-bg.

The output file for filter mode is a text file also with header line.

Example use case on experimetal data

Using real DNA-sequencing data from mouse, we will show how to apply ImmunoPepper to generate all candidate kmers. In this example, we consider two samples: ENCSR000BZG and ERR2130621. We choose ENCSR000BZG as the background sample and ERR2130621 as the foreground sample. They use the same splicegraph but have different expression values and individual (personalized) mutations. Our goal is to generate all kmers unique to ERR2130621.

• Step 1: Use the build mode to generate kmers of the two samples in all four mutation modes:

# reference (ref) mode
immunopepper build --mutation-mode ref --samples ENCSR000BZG ERR2130621 --output-dir ImmunoPepper_usecase_out --splice-path ImmunoPepper_usecase.pickle --ann-path ImmunoPepper_usecase.gtf --ref-path genome1.fa --kmer 9 --count-path ImmunoPepper_usecase.count.hdf5
# germline mode
immunopepper build --mutation-mode germline --samples ENCSR000BZG ERR2130621 --output-dir ImmunoPepper_usecase_out --splice-path ImmunoPepper_usecase.pickle --ann-path ImmunoPepper_usecase.gtf --ref-path genome1.fa --kmer 9 --count-path ImmunoPepper_usecase.count.hdf5 --germline ImmunoPepper_usecase.vcf --somatic ImmunoPepper_usecase.maf
# somatic mode
immunopepper build --mutation_mode somatic --samples ENCSR000BZG ERR2130621 --output-dir ImmunoPepper_usecase_out --splice-path ImmunoPepper_usecase.pickle --ann-path ImmunoPepper_usecase.gtf --ref-path genome1.fa --kmer 9 --count-path ImmunoPepper_usecase.count.hdf5 --germline ImmunoPepper_usecase.vcf --somatic ImmunoPepper_usecase.maf
# germline and somatic mode
immunopepper build --mutation_mode somatic_and_germline --samples ENCSR000BZG ERR2130621 --output-dir ImmunoPepper_usecase_out --splice-path ImmunoPepper_usecase.pickle --ann-path ImmunoPepper_usecase.gtf --ref-path genome1.fa --kmer 9 --count-path ImmunoPepper_usecase.count.hdf5 --germline ImmunoPepper_usecase.vcf --somatic ImmunoPepper_usecase.maf

• Step 2: Create background kmer set from the output of sample ENCSR000BZG. Since there exist no mutations in sample ENCSR000BZG, we only consider its output in reference. In addition, we only consider kmers that have junction expression larger than 0. We can achieve this using the filter mode and get the file ref_mode_background_kmer.tsv. We then use the make_bg mode to create the background kmer file. Since the input is just one file, make_bg simply takes the first column and outputs all unique kmers.

immunopepper filter --output-dir ImmunoPepper_usecase_out --output-file-path ImmunoPepper_usecase_out/ENCSR000BZG/ref_mode_background_kmer.tsv --junction-kmer-tsv-path ImmunoPepper_usecase_out/ENCSR000BZG/ref_junction_kmer.txt --junc_expr
immunopepper make_bg --kmer_files_list ImmunoPepper_usecase_out/ENCSR000BZG/ref_mode_background_kmer.tsv --output-dir ImmunoPepper_usecase_out --output-file-path ImmunoPepper_usecase_out/background_kmer.txt

• Step 3: Remove the background kmers After generating the background kmers in Step 2, we can now subtract those kmers from the kmer sets of sample ERR2130621. We can use diff for this operation:

# contrast ref kmers against background
immunopepper diff --junction-kmer-file  ImmunoPepper_usecase_out/ERR2130621/ref_junction_kmer.txt --bg-file-path ImmunoPepper_usecase_out/background_kmer.txt --output-dir ImmunoPepper_usecase_out --output-file-path ImmunoPepper_usecase_out/ERR2130621/ref_junction_kmer_remove-bg.tsv --remove-bg
# contrast germline kmers against background
immunopepper diff --junction-kmer-file  ImmunoPepper_usecase_out/ERR2130621/germline_junction_kmer.txt --bg-file-path ImmunoPepper_usecase_out/background_kmer.txt --output-dir ImmunoPepper_usecase_out --output-file-path ImmunoPepper_usecase_out/ERR2130621/germline_junction_kmer_remove-bg.tsv --remove-bg
# contrast somatic kmers against background
immunopepper diff --junction-kmer-file  ImmunoPepper_usecase_out/ERR2130621/somatic_junction_kmer.txt --bg-file-path ImmunoPepper_usecase_out/background_kmer.txt --output-dir ImmunoPepper_usecase_out --output-file-path ImmunoPepper_usecase_out/ERR2130621/somatic_junction_kmer_remove-bg.tsv --remove-bg
# contrast somatic/germline kmers against background
immunopepper diff --junction-kmer-file  ImmunoPepper_usecase_out/ERR2130621/somatic_and_germline_junction_kmer.txt --bg-file-path ImmunoPepper_usecase_out/background_kmer.txt --output-dir ImmunoPepper_usecase_out --output-file-path ImmunoPepper_usecase_out/ERR2130621/somatic_and_germline_junction_kmer_remove-bg.tsv --remove-bg

• Step 4: Filter After removing the background kmers in Step 3, we can add more filters to further reduce the number of candidate kmers. For example, we only consider the kmers that have junction expression larger than 0 as well as a segment expression value larger than 2. filter mode provides filters based on segment expression and junction expression, based on a user-provided threshold.

# filter ref kmers
immunopepper filter --output-dir ImmunoPepper_usecase_out --output-file-path ImmunoPepper_usecase_out/ERR2130621/ref_junction_kmer_remove-bg_filter.tsv --junction-kmer-tsv-path ImmunoPepper_usecase_out/ERR2130621/ref_junction_kmer_remove-bg.tsv --cross-junction --seg-expr --seg-expr-thresh 2
# filter germline kmers
immunopepper filter --output-dir ImmunoPepper_usecase_out --output-file-path ImmunoPepper_usecase_out/ERR2130621/germline_junction_kmer_remove-bg_filter.tsv --junction-kmer-tsv-path ImmunoPepper_usecase_out/ERR2130621/germline_junction_kmer_remove-bg.tsv --cross-junction --seg-expr --seg-expr-thresh 2
# filter somatic kmers
immunopepper filter --output-dir ImmunoPepper_usecase_out --output-file-path ImmunoPepper_usecase_out/ERR2130621/somatic_junction_kmer_remove-bg_filter.tsv --junction-kmer-tsv-path ImmunoPepper_usecase_out/ERR2130621/somatic_junction_kmer_remove-bg.tsv --cross-junction --seg-expr --seg-expr-thresh 2
# filter germline/somatic kmers
immunopepper filter --output-dir ImmunoPepper_usecase_out --output-file-path ImmunoPepper_usecase_out/ERR2130621/somatic_and_germline_junction_kmer_remove-bg_filter.tsv --junction-kmer-tsv-path ImmunoPepper_usecase_out/ERR2130621/somatic_and_germline_junction_kmer_remove-bg.tsv --cross-junction --seg-expr --seg-expr-thresh 2

• Step 5: Aggregate We get the unique kmers of sample ERR2130621 in four modes. Now we can aggregate all those kmers.

tail -n +2 ImmunoPepper_usecase_out/ERR2130621/*_junction_kmer_remove-bg_filter.tsv | cat | grep -v "==>" | cut -f1 | sort |uniq | grep . > neo_kmer.txt

Pratical Tips

• ImmunoPepper requires the sample name are exactly the same in the splice count file and mutation file and the given option --samples should be those samples. Please make necessary changes to the input files so that ImmunoPepper can work as expected.

• make_bg, diff and filter mode accept the output files of ImmunoPepper. However, the user can also add other external input files.

• make_bg assumes the input file has a header line, separated with \t and kmers are in the first column.

• diff assumes the foreground kmer file has a header line and that the background kmer file has the format as the output file of make_bg.

• filter assumes the input file has a header line and with three columns seg-expr, junction_expr and is_crossjunction. It's acceptable if some columns are missing but the user should not use corresponding filter rules. Otherwise error will happen.

License

Please see the LICENSE file for more information about license and copyright.