search

adavi4 / SAI-10k-calc

MIT License
8 stars 0 forks source link

readme

SAI-10k-calc

============

1. About

SpliceAI post-processing calculator. Requires two files:

If you use the parser, please cite our manuscript: Canson DM, Davidson AL et al. Bioinformatics 2023 https://doi.org/10.1093/bioinformatics/btad179

The SAI-10k-calc can now optionally accept SpliceAI input files with additional columns containing the REF and ALT scores (the Delta score is the difference between these two scores). Using these scores can lead to improved partial exon deletion and partial intron retention predictions.

We developed the SpliceAI-10k calculator (SAI-10k-calc) to extend use of the SpliceAI tool to predict: the splicing aberration type including: pseudoexonization, intron retention, partial exon deletion, partial intron retention, (multi)exon skipping and increased exon inclusion.

An updated flowchart to reflect these changes can be found here: https://miro.com/app/board/uXjVPD0nK64=/

Authors:
Daffodil Canson - developed the parser
Aimee Davidson - implemented the parser in R and added extra functionality
Olga Kondrashova - implemented the parser in R
QIMR Berghofer Medical Research Institute

2. Requirements

Software Requirements:

3. Usage and Program Options

Step 1. Run download_tx.R script to download and pre-process transcript tables:

Rscript download_tx.R -g example_gene_list.txt --out_refseq example_refseq_tx_hg19.txt --out_tx_spliceai example_spliceai_tx_hg19.txt

Full options are available by running:

Rscript download_tx.R -h

Options: -g FILE, --gene_list=FILE Gene list, tab-separated with two columns - Gene and RefSeq_ID

--out_refseq=FILE
    Save downloaded preprocessed transcript table

--out_tx_spliceai=FILE
    Save preprocessed transcript table for SpliceAI

--out_refseq_full=FILE
    Optional: save downloaded ncbiRefSeq table

--refseq_full=FILE
    Optional: ncbiRefSeq table pre-downloaded from ucsc

--ref=STRING
    Optional: reference genome (hg19 or hg38) [default= hg19]

-h, --help
    Show this help message and exit

Note if you encounter errors with downloading from UCSC using download_tx.R download the UCSC transcript table directly and use the --refseq_full option.

Step 2. Run SpliceAI using the pre-processed transcript table (example_spliceai_tx_hg19.txt) from the first script. Check the example script run_splice_ai.sh.

It is recommended to run SpliceAI with a custom annotation file (pre-processed transcript table) using -A flag, because predictions for specific transcripts may differ from GENCODE V24 canonical transcript predictions (SpliceAI default).

Alternatively, to generate SpliceAI output with the REF and ALT score use this version

Step 3. Run spliceAI_parser.R script using the pre-processed transcript table from Step 1 (example_refseq_tx_hg19.txt) and the SpliceAI output vcf file.

Rscript spliceAI_parser.R -i example_variants.vcf -r example_refseq_tx_hg19.txt -o example_variants_parsed.tsv

Full options are available by running:

Rscript spliceAI_parser.R -h

Options: -i FILE, --in_vcf=FILE Input vcf file direct from SpliceAI

-r FILE, --refseq_table=FILE
    Preprocessed UCSC ncbiRefSeq table (download_tx.R script)

-o FILE, --out_file=FILE
    Output file name [default= out.tsv]

--include
    Option to make use of non-delta scores [default= FALSE]  

--DS_AGDG_MIN_T=NUMERIC
    Delta score (acceptor & donor gain) - minimum [default= 0.02]

--DS_AGDG_MAX_T=NUMERIC
    Delta score (acceptor & donor gain) - maximum [default= 0.2]

--GEX_size_MIN=INT
    Gained exon size range - minimum [default= 25]

--GEX_size_MAX=INT
    Gained exon size range - maximum [default= 500]

--DS_ALDL_MIN_T=NUMERIC
    Delta score (acceptor & donor loss)  - minimum [default= 0.02]

--DS_ALDL_MAX_T=NUMERIC
    Delta score (acceptor & donor loss) - maximum [default= 0.2]

--AG_T=NUMERIC
    Cryptic splice site - acceptor gain [default= 0.2]

--DG_T=NUMERIC
    Cryptic splice site - donor gain [default= 0.2]

--AGDG_T=NUMERIC
    Cryptic splice site - max acceptor or donor gain [default= 0.0],

--DS_GL_T=NUMERIC
    Variant score gain loss difference [default= -0.2]

--ref=STRING
    Optional: reference genome (hg19 or hg38) [default= hg19]

-h, --help
    Show this help message and exit

We have increased the default DS_AGDG_MAX_T threshold to 0.2 to better reflect recommendations by ClinGen SVI Splicing Subgroup (DOI https://doi.org/10.1016/j.ajhg.2023.06.002). However for research purposes we still recommend initially using the previous default threshold of 0.05

Note the parser is currently not fully implemented for variants impacting single exon or single coding exon genes