SVLearn

https://github.com/yangqimeng99/svlearn
Admin
SVLearn is a machine learning-based genotyping tool for structural variation of short reads.
Tokens:7,704
Snippets:72
Trust Score:6
License:MIT
Update:3 months ago
Show doc for...
Tokens:
Raw
### Test SVLearn Installation

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

Navigate to the test directory and execute the test script to verify SVLearn installation.

```bash
cd test
bash test.sh
```

--------------------------------

### Build SVLearn from Tarball

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

Extract the SVLearn tarball, navigate into the directory, download the 'paragraph' binary, unzip it, and then run the installation script. Ensure the conda environment is activated.

```bash
tar xvzf svlearn-0.0.5.tar.gz
cd svlearn-0.0.5
wget https://github.com/Illumina/paragraph/releases/download/v2.4a/paragraph-v2.4a-binary.zip
unzip -q paragraph-v2.4a-binary.zip -d bin/paragraph-v2.4a
bash install.sh
```

--------------------------------

### Build SVLearn from Git Repository

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

Clone the SVLearn repository, change into the directory, download the 'paragraph' binary, unzip it, and execute the installation script. This method requires an activated conda environment.

```bash
git clone https://github.com/yangqimeng99/svlearn.git
cd svlearn
wget https://github.com/Illumina/paragraph/releases/download/v2.4a/paragraph-v2.4a-binary.zip
unzip -q paragraph-v2.4a-binary.zip -d bin/paragraph-v2.4a
bash install.sh
```

--------------------------------

### Create and Activate Conda Environment for SVLearn

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

Use this command to create a new conda environment named 'svlearn' with all the specified Python packages and dependencies. Activate the environment before proceeding with the installation.

```bash
conda create -n svlearn python=3.9 bioconda::pysam=0.22.0 polars=0.20.15 pandas=2.2.1 scikit-learn=1.3.0 bioconda::pyfaidx pyarrow bioconda::pybedtools conda-forge::intervaltree
conda activate svlearn
```

--------------------------------

### Run Cattle 18 Feature Demo

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

Demonstrates SVLearn's genotyping and benchmarking capabilities on a cattle dataset with 18 features. This process is expected to take approximately 5 minutes on a single CPU core with 16GB of memory.

```bash
svlearn genotype \
      -v 02.SV-set_Genomes/cattle_ref_sorted_format_filtered_sv.vcf \
      -m Cattle_18feature_RandomForest_model.joblib \
      -s 04.validation-dataset/cattle/Cattle_SV_Set-sv_feature.tsv \
      -a 04.validation-dataset/cattle/Charolais_30x/BreakPoint_ReadDepth_2Bam_feature.tsv \
      -o Charolais_30_svlearn_genotype.vcf \
      -n Charolais_30

svlearn benchmark \
      -b 04.validation-dataset/cattle/Cattle_SV_Set-Charolais_True_GT.vcf \
      -c Charolais_30_svlearn_genotype.vcf \
      -o Charolais_30_svlearn.bench.tsv \
      -n1 Charolais -n2 Charolais_30
```

--------------------------------

### SVLearn Benchmark Command Line Arguments

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

This snippet shows the usage and available arguments for the `svlearn benchmark` command. Use this to understand the required inputs for comparing VCF files.

```bash
svlearn benchmark
usage: svlearn benchmark [-h] -b file -c file -n1 str -n2 str [-o file]

-h, --help                     show this help message and exit
-b, --base_set file            True sv vcf set
-c, --call_set file            Call sv vcf set
-n1, --base_sample_name str    Sample name in true sv vcf set
-n2, --call_sample_name str    Sample name in call sv vcf set
-o, --out file                 The out of bechmark result, Default: benchmark.result.tsv
```

--------------------------------

### Run Human 24 Feature Demo

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

Demonstrates SVLearn's genotyping and benchmarking capabilities on a human dataset with 24 features. This process is expected to take approximately 5 minutes on a single CPU core with 16GB of memory.

```bash
svlearn genotype \
      -v 02.SV-set_Genomes/human_ref_sorted_format_filtered_sv.vcf \
      -m Human_30x_24feature_RandomForest_model.joblib \
      -s 04.validation-dataset/human/Human_SV_Set-sv_feature.tsv \
      -a 04.validation-dataset/human/HG002_30x/BreakPoint_ReadDepth_2Bam_feature.tsv \
      -p 04.validation-dataset/human/HG002_30x/para_feature.tsv \
      -o HG002_30_svlearn_genotype.vcf \
      -n HG002_30

svlearn benchmark \
      -b 04.validation-dataset/human/Human_SV_Set-HG002_True_GT.vcf \
      -c HG002_30_svlearn_genotype.vcf \
      -o HG002_30_svlearn.bench.tsv \
      -n1 HG002 -n2 HG002_30
```

--------------------------------

### Prepare Training Label File

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

Format for the training label file, which should be tab-separated. It includes SV IDs and their ground truth labels.

```text
sv_id	GT_true
INS.0	0/1
DEL.0	0/0
INS.1	1/1
.....	.....
```

--------------------------------

### Create Genotype Directory and Navigate

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

This command creates a new directory for genotyping results and changes the current directory into it. This is a prerequisite for running the genotyping commands.

```bash
mkdir 06.genotype;cd 06.genotype
```

--------------------------------

### Prepare Alt Genome with Filtering

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

Command to prepare an alternative genome by filtering overlapping SVs. This is the default behavior and prepares files for subsequent analysis.

```bash
svlearn prepareAlt --ref_fasta ref.fasta --ref_sv_vcf sv.vcf --out 01.prepareAlt_output
```

--------------------------------

### Navigate Back to Parent Directory

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

After completing the genotyping process, this command changes the directory back to the parent directory. This is a common practice to return to the main project structure.

```bash
cd ..
```

--------------------------------

### Create and Activate Mamba Environment for SVLearn

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

This command uses mamba to create a new environment named 'svlearn' with the specified packages. Mamba is generally faster than conda for environment creation. Activate the environment after creation.

```bash
mamba create -n svlearn python=3.9 bioconda::pysam=0.22.0 polars=0.20.15 pandas=2.2.1 scikit-learn=1.3.0 bioconda::pyfaidx pyarrow bioconda::pybedtools conda-forge::intervaltree
mamba activate svlearn
```

--------------------------------

### Generate Training Matrix with 18 Features

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

Uses csvtk to join ground truth labels with feature files to create an 18-feature training matrix. This process is repeated for two sample sets.

```bash
# 18 features
csvtk join -t -f 'sv_id' training_sample1.true.gt.tsv sv_feature.tsv training_sample1.BreakPoint_ReadDepth_2Bam_feature.tsv > training_sample1.18feature.training.tsv
csvtk join -t -f 'sv_id' training_sample2.true.gt.tsv sv_feature.tsv training_sample2.BreakPoint_ReadDepth_2Bam_feature.tsv > training_sample2.18feature.training.tsv

csvtk concat training_sample1.18feature.training.tsv training_sample2.18feature.training.tsv > sample1_2.18feature.training.tsv
```

--------------------------------

### Generate Training Matrix with 24 Features

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

Uses csvtk to join ground truth labels with feature files, including additional 'para_feature.tsv', to create a 24-feature training matrix. This process is repeated for two sample sets.

```bash
# 24 features
csvtk join -t -f 'sv_id' training_sample1.true.gt.tsv sv_feature.tsv training_sample1.BreakPoint_ReadDepth_2Bam_feature.tsv training_sample1.para_feature.tsv > training_sample1.24feature.training.tsv
csvtk join -t -f 'sv_id' training_sample2.true.gt.tsv sv_feature.tsv training_sample2.BreakPoint_ReadDepth_2Bam_feature.tsv training_sample2.para_feature.tsv > training_sample2.24feature.training.tsv

csvtk concat training_sample1.24feature.training.tsv training_sample2.24feature.training.tsv > sample1_2.24feature.training.tsv
```

--------------------------------

### Cattle Demo Benchmark Results

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

Displays the performance metrics for the cattle SVLearn genotyping results compared to the truth set. Metrics like precision_GT and recall_GT focus on distinguishing heterozygous and homozygous genotypes.

```text
sv_set_number                 121435
genotyped_sv_number           120151
genotype_rate                 0.9894
accuracy_genotyped_sv_number  115362
precision_GT                  0.8619
recall_GT                     0.7924
f1_GT                         0.8257
precision                     0.9104
recall                        0.837
f1                            0.8721
conc_00                       0.9854
conc_01                       0.7866
conc_11                       0.8621
wgc                           0.878
```

--------------------------------

### Run svlearn runParagraph

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

Extracts paragraph features for SV sites from BAM files using the Paragraph tool. Requires reference and alternative FASTA, VCF, and BAM files. Outputs a TSV file with paragraph features.

```bash
mkdir 05.para_feature;cd 05.para_feature

svlearn runParagraph \
        --ref_fasta ref.fasta \
        --alt_fasta alt.fasta \
        --ref_sv_vcf ../01.prepareAlt_output/ref_sorted_format_filtered_sv.vcf \
        --alt_sv_vcf ../01.prepareAlt_output/alt_sorted_format_filtered_sv.vcf \
        --ref_bam ../03.mapping/01.RefGenome/sample1.dedup.sort.bam \
        --alt_bam ../03.mapping/02.AltGenome/sample1_alt.dedup.sort.bam \
        --threads 20 \
        --out sample1

cd ..
```

--------------------------------

### Calculating Depth and Preparing Manifest for Paragraph

Source: https://github.com/yangqimeng99/svlearn/wiki/Compare-with-other-tools

Calculates sequencing depth for a sample using its BAM file and reference genome, then formats this information into a manifest file required by Paragraph for genotyping.

```bash
# depth
mkdir -p ./depth ./manifest
idxdepth -b sample1.dedup.sort.bam -r ref.fasta -o ./depth/sample1.json --threads 16
echo -e "id\tpath\tdepth\tread length" > ./manifest/sample1.manifest.txt
Depth_ReadsLength=$(awk '{match($0,/,"depth":([0-9]+\.[0-9]+)/,a);match($0, /"read_length":([0-9]+)/,b);print a[1]"\t"b[1]}' depth/sample1.json)
echo -e "sample1\tsample1.dedup.sort.bam\t${Depth_ReadsLength}" >> ./manifest/sample1.manifest.txt
```

--------------------------------

### Genotype and Benchmark SVs

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

Commands to genotype structural variants using a trained model and then benchmark the results against a truth set. Ensure input file paths are correct for your environment.

```bash
svlearn genotype \
      -v 02.SV-set_Genomes/sheep_ref_sorted_format_filtered_sv.vcf \
      -m Sheep_18feature_RandomForest_model.joblib \
      -s 04.validation-dataset/sheep/Sheep_SV_Set-sv_feature.tsv \
      -a 04.validation-dataset/sheep/Romanov_30x/BreakPoint_ReadDepth_2Bam_feature.tsv \
      -o Romanov_30_svlearn_genotype.vcf \
      -n Romanov_30

svlearn benchmark \
      -b 04.validation-dataset/sheep/Sheep_SV_Set-Romanov_True_GT.vcf \
      -c Romanov_30_svlearn_genotype.vcf \
      -o Romanov_30_svlearn.bench.tsv \
      -n1 Roman and -n2 Romanov_30
```

--------------------------------

### Train Model with 24 Features

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

Trains a RandomForest model using the generated 24-feature training set, including 'paragraph' features. Specify the training set, model type, output file name, and number of threads.

```bash
# 24 features
svlearn trainingModel \
      --train_set sample1_2.24feature.training.tsv \
      --other_feature paragraph \
      --train_model RandomForest \
      --out RandomForest.24feature.model
      --threads 64
```

--------------------------------

### Run svlearn Genotype with 18-Feature Model

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

This command runs the svlearn genotype function using an 18-feature model. It requires reference SV VCF, a trained model, SV features, and alignment features. Use this when paragraph features have not been extracted.

```bash
svlearn genotype \
        --ref_sv_vcf ../01.prepareAlt_output/ref_sorted_format_filtered_sv.vcf \
        --model model.joblib \
        --sv_feature ../02.SV.feature/sv_feature.tsv \
        --align_feature ../04.align_feature/sample1/BreakPoint_ReadDepth_2Bam_feature.tsv \
        --name sample1 \
        --out sample1_svlearn_18feature_genotype.vcf
```

--------------------------------

### Train Model with 18 Features

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

Trains a RandomForest model using the generated 18-feature training set. Specify the training set, model type, output file name, and number of threads.

```bash
# 18 features
svlearn trainingModel \
      --train_set sample1_2.18feature.training.tsv \
      --train_model RandomForest \
      --out RandomForest.18feature.model
      --threads 64
```

--------------------------------

### Run svlearn Genotype with 24-Feature Model

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

This command runs the svlearn genotype function using a 24-feature model, which includes paragraph features. It requires all inputs from the 18-feature model plus the paragraph features. Use this when paragraph features have been extracted.

```bash
svlearn genotype \
        --ref_sv_vcf ../01.prepareAlt_output/ref_sorted_format_filtered_sv.vcf \
        --model model.joblib \
        --sv_feature ../02.SV.feature/sv_feature.tsv \
        --align_feature ../04.align_feature/sample1/BreakPoint_ReadDepth_2Bam_feature.tsv \
        --paragraph_feature ../05.para_feature/sample1/para_feature.tsv \
        --name sample1 \
        --out sample1_svlearn_24feature_genotype.vcf
```

--------------------------------

### Prepare Alt Genome without Filtering Overlaps

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

Command to prepare an alternative genome while retaining overlapping SVs. Use this option if you do not want to filter out overlapping SVs.

```bash
svlearn prepareAlt --ref_fasta ref.fasta --ref_sv_vcf sv.vcf --out 01.prepareAlt_output --no-filter-overlaps
```

--------------------------------

### Run svlearn alignFeature

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

Extracts alignment features for SV sites from BAM files. Requires reference and alternative FASTA, VCF, BED, and BAM files. Outputs TSV files containing alignment features.

```bash
mkdir 04.align_feature;cd 04.align_feature

svlearn alignFeature \
        --ref_fasta ref.fasta \
        --alt_fasta alt.fasta \
        --ref_sv_vcf ../01.prepareAlt_output/ref_sorted_format_filtered_sv.vcf \
        --alt_sv_bed ../01.prepareAlt_output/alt_sorted_format_filtered_sv.bed \
        --ref_bam ../03.mapping/01.RefGenome/sample1.dedup.sort.bam \
        --alt_bam ../03.mapping/02.AltGenome/sample1_alt.dedup.sort.bam \
        --threads 1 \
        --out sample1

cd ..
```

--------------------------------

### Prepare Alternative Insertion VCF

Source: https://github.com/yangqimeng99/svlearn/wiki/Compare-with-other-tools

Filters the alternative VCF for deletions (treated as insertions) and adds CIPOS/CIEND tags. This step aligns the alternative VCF to the reference VCF's format for SVTyper processing.

```bash
bcftools view -i 'SVTYPE="DEL"' alt_sorted_format_filtered_sv.vcf | awk '{if($0 ~ /^#/) print $0; else print $0 ";CIPOS=-100,100;CIEND=-100,100"}' > alt_ins_ann.vcf
```

--------------------------------

### Extract SV Features with svlearn

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

Integrates features extracted by RepeatMasker, trf, GenMap, and BISER for SV analysis. Ensure all input files are correctly specified.

```bash
mkdir 02.SV.feature; cd 02.SV.feature

# 01.RepeatMasker, Please adjust the `-pa` and `-species` parameters based on the actual circumstances.
RepeatMasker -pa 64 -engine ncbi -species human -xsmall -s -no_is -cutoff 255 -frag 20000 -dir ./ -gff ref.fasta
RepeatMasker -pa 64 -engine ncbi -species human -xsmall -s -no_is -cutoff 255 -frag 20000 -dir ./ -gff alt.fasta

# 02.trf
trf ref.fasta 2 7 7 80 10 50 500 -f -d -h
trf alt.fasta 2 7 7 80 10 50 500 -f -d -h
python /software/svlearn/script/TRF2GFF.py -d ref.fasta.2.7.7.80.10.50.500.dat -o ref.trf.gff
python /software/svlearn/script/TRF2GFF.py -d alt.fasta.2.7.7.80.10.50.500.dat -o alt.trf.gff

# 03.GenMap
bash /software/svlearn/script/genmap.sh ref.fasta ref.fasta.genmapK50E1
bash /software/svlearn/script/genmap.sh alt.fasta alt.fasta.genmapK50E1

# 04.BISER, ref.fasta.masked and alt.fasta.masked are the softmasked FASTA files obtained from the output of 01.RepeatMasker
biser -o ref.fasta.masked.out -t 2 --gc-heap 32G ref.fasta.masked
biser -o alt.fasta.masked.out -t 2 --gc-heap 32G alt.fasta.masked

# 05.svlearn svFeature
svlearn svFeature \
        --ref_sv_vcf ../01.prepareAlt_output/ref_sorted_format_filtered_sv.vcf \
        --alt_sv_bed ../01.prepareAlt_output/alt_sorted_format_filtered_sv.bed \
        --ref_rm ref.fasta.out \
        --alt_rm alt.fasta.out \
        --ref_trf ref.trf.gff \
        --alt_trf alt.trf.gff \
        --ref_genmap ref.fasta.genmapK50E1.txt \
        --alt_genmap alt.fasta.genmapK50E1.txt \
        --ref_biser ref.fasta.masked.out \
        --alt_biser alt.fasta.masked.out \
        --out sv_feature.tsv

cd ..
```

--------------------------------

### Run SVTyper for Deletions and Insertions

Source: https://github.com/yangqimeng99/svlearn/wiki/Compare-with-other-tools

Executes SVTyper to genotype deletions using the reference BAM and insertions using the alternative BAM. This step generates VCF files with SVTyper's genotype calls for both deletion and insertion variants.

```bash
svtyper -i ref_del_ann.vcf -B sample1.dedup.sort.bam -l sample1_ref.bam.json > sample1_svtyper_del.vcf
```

```bash
svtyper -i alt_ins_ann.vcf -B sample1_alt.dedup.sort.bam -l sample1_alt.bam.json > sample1_svtyper_ins.vcf
```

--------------------------------

### Human Demo Benchmark Results

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

Displays the performance metrics for the human SVLearn genotyping results compared to the truth set. Metrics like precision_GT and recall_GT focus on distinguishing heterozygous and homozygous genotypes.

```text
sv_set_number                 38613
genotyped_sv_number           37021
genotype_rate                 0.9588
accuracy_genotyped_sv_number  32956
precision_GT                  0.8292
recall_GT                     0.7585
f1_GT                         0.7922
precision                     0.9209
recall                        0.8424
f1                            0.8799
conc_00                       0.95
conc_01                       0.7534
conc_11                       0.8575
wgc                           0.8537
```

--------------------------------

### Prepare Reference Deletion VCF

Source: https://github.com/yangqimeng99/svlearn/wiki/Compare-with-other-tools

Filters the reference VCF for deletions and adds CIPOS/CIEND tags. This prepares the VCF for SVTyper by ensuring only deletion variants are considered and standard confidence intervals are applied.

```bash
bcftools view -i 'SVTYPE="DEL"' ref_sorted_format_filtered_sv.vcf | bcftools +fill-tags - -- -t END | awk '{if($0 ~ /^#/) print $0; else print $0 ";CIPOS=-100,100;CIEND=-100,100"}' > ref_del_ann.vcf
```

--------------------------------

### Generating Bloom Filter for BayesTyper

Source: https://github.com/yangqimeng99/svlearn/wiki/Compare-with-other-tools

Creates a Bloom filter using BayesTyper tools, which is used for efficient variant querying. This step requires the k-mer count output from KMC3.

```bash
bayesTyperTools makeBloom -k sample1 -p 16
```

--------------------------------

### KMC3 K-mer Counting for BayesTyper

Source: https://github.com/yangqimeng99/svlearn/wiki/Compare-with-other-tools

Counts k-mers from a sample's BAM file using KMC3, which is a preliminary step for BayesTyper. The output is stored in a specified temporary directory.

```bash
#before bayestyper, using KMC3 to Count k-mers.
mkdir -p ./kmc_tmp
kmc -k55 -ci1 -fbam sample1.dedup.sort.bam sample1 ./kmc_tmp
```

--------------------------------

### Genotyping with Paragraph

Source: https://github.com/yangqimeng99/svlearn/wiki/Compare-with-other-tools

Performs structural variant genotyping using the Paragraph tool. It requires a manifest file with depth information and a set of SVs in VCF format.

```bash
Depth=`cut -f3 ./manifest/sample1.manifest.txt | tail -1`
maxDepth=`echo 20*${Depth}/1 | bc`

#genotype
mkdir -p tmp
multigrmpy.py \
    -i ref_sorted_format_filtered_sv.vcf \
    -M ${maxDepth} \
    -m ./manifest/sample1.manifest.txt \
    -r ref.fasta \
    -t 16 -o sample1 --scratch-dir tmp
```

--------------------------------

### Genotyping with BayesTyper

Source: https://github.com/yangqimeng99/svlearn/wiki/Compare-with-other-tools

Performs variant genotyping for each unit using BayesTyper. This command requires variant cluster data, sample information, reference genomes, and output parameters.

```bash
# genotyping (${i}=1、2...n)
bayesTyper genotype -r 10 -v bayestyper_unit_${i}/variant_clusters.bin -c bayestyper_cluster_data -s sample.tsv -g ref.canon.fasta -d ref.decoy.fasta -o bayestyper_unit_${i}/bayestyper -z -p 16
```

--------------------------------

### Construct Final Genotype VCF

Source: https://github.com/yangqimeng99/svlearn/wiki/Compare-with-other-tools

Builds the final VCF file containing combined genotype information. This includes header information and joins the genotype data with variant details from the reference VCF.

```bash
bcftools view ref.vcf | grep '##' > sample1_svtyper_sv.gt.vcf
```

```bash
echo -e "##FORMAT=<ID=GT,Number=1,Type=String,Description=\"Genotype\">" >> sample1_svtyper_sv.gt.vcf
```

```bash
echo -e "#CHROM\tPOS\tID\tREF\tALT\tQUAL\tFILTER\tINFO\tFORMAT\tsvtyper_GT" >> sample1_svtyper_sv.gt.vcf
```

```bash
bcftools view ref_sorted_format_filtered_sv.vcf | grep -v '#' | awk '{print $0"\tGT"}' | csvtk join -t -H -f '3;1' - sample1_svtyper_sv.gt.tsv >> sample1_svtyper_sv.gt.vcf
```

--------------------------------

### Genotyping Structural Variants with GraphTyper2

Source: https://github.com/yangqimeng99/svlearn/wiki/Compare-with-other-tools

Executes GraphTyper2 for genotyping structural variants. It requires the reference genome, an SV VCF file, a BAM file, and a file specifying genomic regions for genotyping.

```bash
graphtyper genotype_sv ref.fasta ref_sorted_format_filtered_sv.vcf.gz \
    --sam=sample1.dedup.sort.bam \
    --region_file=genotyping_sv.region \
    --output=sample1 \
    --threads=16

bcftools concat -a ./sample1/*/*.vcf.gz -O z -o sample1_graphtyper.vcf.gz
```

--------------------------------

### Indexing Reference Genome with BWA-MEM2

Source: https://github.com/yangqimeng99/svlearn/wiki/Compare-with-other-tools

Indexes the reference genome FASTA file for use with BWA-MEM2 aligner. This is a prerequisite for mapping reads.

```bash
bwa-mem2.avx512bw index ref.fasta
```

--------------------------------

### Extract and Format Genotypes

Source: https://github.com/yangqimeng99/svlearn/wiki/Compare-with-other-tools

Queries the genotype information from the SVTyper output VCFs and combines them into a single TSV file. It also inverts genotypes for insertions to match the reference genome's perspective.

```bash
bcftools query -f '%ID\t[%GT]\n' sample1_svtyper_del.vcf > sample1_svtyper_sv.gt.tsv
```

```bash
bcftools query -f '%ID\t[%GT]\n' sample1_svtyper_ins.vcf | awk '{if ($2=="0/0") {print $1"\t1/1"} else if ($2=="1/1") {print $1"\t0/0"} else {print $0}}' >> sample1_svtyper_sv.gt.tsv
```

--------------------------------

### Map Reads to Reference Genome with BWA-MEM2

Source: https://github.com/yangqimeng99/svlearn/blob/main/README.md

Performs read mapping of sample short reads to the reference FASTA file using BWA-MEM2 and a helper script for deduplication and sorting. This process is repeated for the alternative genome.

```bash
# Reads mapping, only one sample is shown as an example, but multiple samples can be mapped in parallel
cd 01.RefGenome
bash /software/svlearn/script/bwa_dedup.sh sample1_R1.fastq.gz sample1_R2.fastq.gz ref.fasta sample1
cd ../02.AltGenome
bash /software/svlearn/script/bwa_dedup.sh sample1_R1.fastq.gz sample1_R2.fastq.gz alt.fasta sample1_alt

cd ../..
```

--------------------------------

### Variant Clustering with BayesTyper

Source: https://github.com/yangqimeng99/svlearn/wiki/Compare-with-other-tools

Identifies variant clusters using BayesTyper. This command requires the reference genome (canonical and decoy), an SV VCF file, a sample TSV file, and the number of threads.

```bash
#Identifying variant clusters
bayesTyper cluster -r 10 -v ref_sorted_format_filtered_sv.vcf.gz -s sample.tsv -g ref.canon.fasta -d ref.decoy.fasta -p 16
```

--------------------------------

### Mapping Reads with BWA-MEM2 and Deduplication

Source: https://github.com/yangqimeng99/svlearn/wiki/Compare-with-other-tools

Maps short reads to the reference genome using BWA-MEM2 and performs deduplication. This script generates a sorted and deduplicated BAM file.

```bash
# Reads mapping
bash /software/svlearn/script/bwa_dedup.sh sample1_R1.fastq.gz sample1_R2.fastq.gz ref.fasta sample1

# output BAM: sample1.dedup.sort.bam
```

--------------------------------

### Merging BayesTyper Genotyping Results

Source: https://github.com/yangqimeng99/svlearn/wiki/Compare-with-other-tools

Concatenates the VCF files generated by BayesTyper for each unit into a single VCF file. This command uses bcftools to merge the results.

```bash
#merge
bcftools concat -a -o sample1_bayestyper.vcf.gz bayestyper_unit_*/bayestyper.vcf.gz
```
Tokens:
Raw