注释的话 从这里可以下载 gff文件 用 snpeff或者 ANNOVAR 进行注释

或者
perl /home/admin/software/annovar/table_annovar.pl $INPUT
/home/admin/software/annovar/humandb -buildver hg19
-out $OUT_DIR/$PREFIX.annovar -remove
-protocol refGene,clinvar_20170905
-operation g,f
-nastring .
-vcfinput

gatk SelectVariants
-R ref.fa
-V raw_variants_recal.vcf
-selectType SNP
-o raw_snps_recal.vcf
gatk SelectVariants
-R ref.fa
-V raw_variants.vcf
-selectType INDEL
-o raw_indels_recal.vcf

gatk VariantFiltration \
-R ref.fa
-V raw_snps_recal.vcf
-O filtered_snps_final.vcf
-filter-name "QD_filter" -filter "QD < 2.0"
-filter-name "FS_filter" -filter "FS > 60.0"
-filter-name "MQ_filter" -filter "MQ < 40.0"
-filter-name "SOR_filter" -filter "SOR > 4.0"
-filter-name "MQRankSum_filter" -filter "MQRankSum < -12.5"
-filter-name "ReadPosRankSum_filter" -filter "ReadPosRankSum < -8.0"

gatk VariantFiltration \
-R ref.fa
-V raw_indels_recal.fa
-O filtered_indels_final.vcf
-filter-name "QD_filter" -filter "QD < 2.0"
-filter-name "FS_filter" -filter "FS > 200.0"
-filter-name "SOR_filter" -filter "SOR > 10.0"

gatk ApplyBQSR
-R ref.fa
-I sorted_dedup_reads.bam
-bqsr recal_data.table
-O recal_reads.bam \

第二轮（可选）
gatk BaseRecalibrator
-R ref.fa
-I recal_reads.bam
--known-sites bqsr_snps.vcf
--known-sites bqsr_indels.vcf
-O post_recal_data.table
分析数据：
gatk AnalyzeCovariates \
-before recal_data.table
-after post_recal_data.table
-plots recalibration_plots.pdf

gatk --java-options "-Xmx8G" Mutect2
-R /home/admin/database/reference/hg19/ucsc.hg19.fasta
-I ../out/normal_recal.bam
-I ../out/cancer_recal.bam
-tumor cancer
-normal normal
--germline-resource af-only-gnomad.raw.sites.hg19.vcf.gz
-L ../in/Illumina.bed
-O ../out/somatic.vcf

gatk --java-options "-Xmx8G" FilterMutectCalls
-V ../out/somatic.vcf
-O ../out/somatic_filtered.vcf.gz

获取SNP和Indels
gatk SelectVariants
-R ref.fa
-V raw_variants.vcf
-selectType SNP
-o raw_snps.vcf
gatk SelectVariants
-R ref.fa
-V raw_variants.vcf
-selectType INDEL
-o raw_indels.vcf

过滤SNP和Indels
gatk VariantFiltration
-R ref.fa
-V raw_snps.vcf
-O filtered_snps.vcf
-filter-name "QD_filter" -filter "QD < 2.0"
-filter-name "FS_filter" -filter "FS > 60.0"
-filter-name "MQ_filter" -filter "MQ < 40.0"
-filter-name "SOR_filter" -filter "SOR > 4.0"
-filter-name "MQRankSum_filter" -filter "MQRankSum < -12.5"
-filter-name "ReadPosRankSum_filter" -filter "ReadPosRankSum < -8.0"

gatk VariantFiltration
-R ref.fa
-V raw_indels.vcf
-O filtered_indels.vcf
-filter-name "QD_filter" -filter "QD < 2.0"
-filter-name "FS_filter" -filter "FS > 200.0"
-filter-name "SOR_filter" -filter "SOR > 10.0"

gatk SelectVariants
--exclude-filtered
-V filtered_snps.vcf
-O bqsr_snps.vcf
gatk SelectVariants
--exclude-filtered
-V filtered_indels.vcf
-O bqsr_indels.vcf

Mark Duplicates -> RBQS -> IndelRealigning by Mutect2

which was below in previous version of GATK

Mark Duplicates -> IndelRealigner -> RBQS -> Mutect2

RBQS without INDEL realigning wouldn’t be affected by false alignments ?

/home/admin/software/gatk-4.1.0.0/gatk BaseRecalibrator
-I my_reads.bam
-R reference.fasta
--known-sites dbsnp_138.hg19.vcf
--known-sites Mills_and_1000G_gold_standard.indels.hg19.sites.vcf
--known-sites 1000G_phase1.indels.hg19.sites.vcf
--known-sites 1000G_phase1.snps.high_confidence.hg19.sites.vcf
-L ../in/Illumina.bed
-O recal_data.table

/home/admin/software/gatk-4.1.0.0/gatk ApplyBQSR
-R reference.fasta
-I input.bam
--bqsr-recal-file recal_data.table
-L ../in/Illumina.bed
-O output.bam

对于非人类的物种怎么处理？
https://gencore.bio.nyu.edu/variant-calling-pipeline-gatk4/
博得推荐的方法是bootstrapping 　即先call SNP　然后用这些数据再做BQSR
Base Quality Score Recalibration (BQSR) is an important step for accurate variant detection that aims to minimize the effect of technical variation on base quality scores (measured as Phred scores). As with the original pipeline (link), this pipeline assumes that a ‘gold standard’ set of SNPS and indels are not available for BQSR. In the absence of a gold standard the pipeline performs an initial step detecting variants without performing BQSR, and then uses the identified SNPs as input for BQSR before calling variants again. This process is referred to as bootstrapping and is the procedure recommended by the Broad Institute’s best practices for variant discovery analysis when a gold standard is not available.