Variant Annotation and Filtration with External Databases enhance the analysis of genetic variations by leveraging external databases to provide additional information about variants. This process aids in filtering and prioritizing variants based on various annotations.

Gene

Gene feature Annotation & Filtration

Gene feature annotation is used to identify which gene or transcript is affected by a variant and what functional role it has on known genes. We now support two gene definition systems: RefSeq genes (refgene) and GENCODE genes (gencode).

Options

The tutorial command is:

java -Xmx8g -jar ../kggsum.jar \
   annot \
   --ref-gty-file ./1kg_hg19_eur_chr1.vcf.gz \
              refG=hg19 \
   --sum-file ./scz_gwas_eur_chr1.tsv.gz \
              cp12Cols=CHR,BP,A1,A2 \
              pbsCols=P,OR,SE \
              betaType=2 \
              prevalence=0.01 \
   --gene-model-database refgene \
   --gene-model-database gencode \
   --threads 18 \
   --output ./ta1

KGGA has 18 number codes for the gene features after annotation.

Output

The gene feature annotation results are saved in OutputVariants2TSVTask/variants.hg38.tsv.gz. There are two relevant columns in the file:

Clump

Prune and Clump Variants Based on P-Values and Linkage Disequilibrium (LD)

In KGGSum, the "Clump" feature enables users to prune and clump Genome-Wide Association Study (GWAS) variants using p-values from summary statistics and linkage disequilibrium (LD) calculated from a reference population. This process produces a refined set of independent, statistically significant, and harmonized variants, ready for downstream applications such as Mendelian Randomization (MR) analysis with third-party tools or platforms.

LD Clumping Methodology

The LD clumping method filters out variants within a user-defined physical distance (set by windowKb) if their LD exceeds a specified threshold (controlled by r2Cut). Variant removal follows a prioritized, three-step approach to retain the most informative variants:

1. Variants with Larger P-Values Variants with less significant p-values (i.e., larger values) are prioritized for removal. The significance threshold is determined by the family-wise error rate (set by pCut) and adjusted using multiple testing correction methods: "fixed" (no correction), "bonf" (Bonferroni), "bhfdr" (Benjamini-Hochberg FDR), "byfdr" (Benjamini-Yekutieli FDR), or "localfdr" (local FDR).
   • When multiple GWAS datasets are provided, the p-value to use is specified by pIndex; otherwise, the p-value from the first dataset is used by default.
   • Among variants with p-values below the cutoff, those with the smallest sum of -log(p) across datasets or missing p-values in any dataset are prioritized for removal.
2. Variants with More LD Connections Variants in LD with a greater number of other variants within the windowKb distance are prioritized for removal to reduce redundancy.
   • For example, consider three variants: A, B, and C, with LD r² values of 0.8 (A-B), 0.8 (B-C), and 0.64 (A-C). Variant B, which is in LD with both A and C, is removed, while A and C are retained. This preserves variants with fewer LD dependencies.
3. Variants with Lower Minor Allele Frequency (MAF) If the first two criteria do not differentiate between variants (e.g., equal LD connections and similar p-values), the variant with the lower MAF is removed. Retaining variants with higher MAF ensures greater informativeness, as they are typically more robust for statistical analysis.

This hierarchical approach maximizes the retention of variants that are:

• Less redundant (fewer LD connections),
• More significant (smaller p-values),
• More informative (higher MAF).

Options

The tutorial command is:

java -Xmx8g -jar ../kggsum.jar \
   annot \
   --ld-clump r2Cut=0.1 \
              windowKb=10000 \
              pIndex=1 \
              pCut=0.05 \              
              pCorrect=bhfdr \
   --ref-gty-file ./1kg_hg19_eur_chr1.vcf.gz \
              refG=hg19 \
   --sum-file ./smoking_chr1.tsv.gz \
              cp12Cols=CHR,POS,A1,A2 \
              pbsCols=Pval,Beta,SE \
              prevalence=0.05   \
              betaType=1   \
   --sum-file ./scz_gwas_eur_chr1.tsv.gz \
              cp12Cols=CHR,BP,A1,A2 \
              pbsCols=P,OR,SE \
              betaType=2 \
              prevalence=0.01 \     
   --threads 18 \
   --output ./ta2

This command clumps variants from two GWAS datasets (smoking and schizophrenia), prioritizing the smoking dataset’s p-values. It uses a 10 Mb window and an LD threshold of 0.1, retaining independent, significant variants after FDR correction.

Function

Functional score annotation at variants

In KGGSum, the GWAS variants can also be annotated with multiple genomic features. Three databases are available: gnomAD for allele frequency annotation, CADD for variant function annotation, and ClinVar for disease linkage annotation. Note that the annotation datasets should be downloaded from an independent resource domain of KGGA.

Options

The tutorial command is:

java -Xmx8g -jar ../kggsum.jar \
   annot \
   --ref-gty-file ./1kg_hg19_eur_chr1.vcf.gz \
              refG=hg19 \
   --sum-file ./scz_gwas_eur_chr1.tsv.gz \
              cp12Cols=CHR,BP,A1,A2 \
              pbsCols=P,OR,SE \
              betaType=2 \
              prevalence=0.01 \
   --variant-annotation-database cadd \
              field=Epigenetics::EncodeDNase-max,Epigenetics::EncodeDNase-sum,ProteinFunction::CADD_PHRED \
   --threads 18 \
   --output ./ta2

Additional epigenetic resources from third-party databases

To facilitate the convenient use of more resources, KGGSum provides an approach that allows users to specify customized third-party resources for annotation. For example, by setting the file name documented in EpiMap, KGGSum can directly download epigenetic marker resources from the EpiGenome public domain, specifically from the EpiMap Repository.

Frequency

Allele Frequency Annotation & Filtration

Allele frequency annotation allows users to incorporate population-level allele frequency information into the analysis of genetic variants. Click to view the provided allele frequency annotation databases.

Options

The tutorial command is:

java -Xmx4g -jar ../kggsum.jar \
   annot \
   --ref-gty-file ./1kg_hg19_eur_chr1.vcf.gz \
              refG=hg19 \
   --sum-file ./scz_gwas_eur_chr1.tsv.gz \
              cp12Cols=CHR,BP,A1,A2 \
              pbsCols=P,OR,SE \
              betaType=2 \
              prevalence=0.01 \       
     --freq-database gnomad \
              field=gnomAD_joint::ALL,gnomAD_joint::NFE \
   --threads 18 \
   --output ./ta3

Once the reference databases for allele frequency annotation have been properly configured, you can effectively filter variants by examining their allele frequencies within the reference population.