snp-calling.md

GATK SNP calling

all scripts are located in snp-calling directory.

1. Input files were downloaded from CyVerse using getFiles.sh script. The files that were large (ran without barcodes), were split to smaller chunks for quick processing using split-fastq.sh script.
2. The genome file was downloaded from CyVerse (B73.v5) and processed using gatk-prepare-reference.sh to create all the necessary files for running GATK pipeline.
3. The fastq files were mapped to B73.v5 and processed using process-fastq.sh script. Briefly, this script:
   • converts unmapped fastq to bam FastqToSam
   • runs Picard MarkIlluminaAdapters
   • converts bam back to fastq SamToFastq
   • Maps fastq files to B73.v5 using bwa mem and converts to bam file using samtools
   • merged unmapped reads with mapped reads using MergeBamAlignment
   • runs picard's MarkDuplicates to mark optical duplicates.
4. As a final step of processing, using run-add-readgroups.sh correct read groups were added to the bam files and indexed.
5. GATK was run on 1Mb intervals, using the script gatkcmds-round-1.sh and the intervals file B73.PLATINUM.pseudomolecules-v1_1mb_coords.bed, the commands were generated and was run on the cluster creating slurm job submission script using GNU parallel.
6. Once the VCF files were generated (2,813 total), they were gathered and processed to filter and retain very high quality SNPs only, using the script gatk-process.sh
7. The bam files were recalibrated using the filtered first round SNP files using gatk-bsqr.sh
8. Using the recalibrated BAM files, GATK was ran again on 1Mb intervals, using the script gatkcmds-round-2.sh and the intervals file B73.PLATINUM.pseudomolecules-v1_1mb_coords.bed, the commands were generated and was run on the cluster creating slurm job submission script using GNU parallel.
9. The final files were filtered again using the gatk-process.sh script again.
10. The final files were uploaded to CyVerse