Reads are initially mapped to the reference genome using a fast algorithm, and these initial mappings are both expanded and refined by a form of local de novo assembly, which is applied to all regions of the genome that appear to contain variation (SNPs, indels, and block substitutions) based on these initial mappings. The de novo assembly fully leverages mate-pair information, allowing reads to be recruited into variant calling with higher sensitivity than genome-wide mapping methods alone typically provide. Assemblies are diploid, and we produce two separate result sequences for each locus in diploid regions (exceptions: mitochondria are assembled as haploid and for males the nonpseudoautosomal regions in the sex chromosomes are assembled as haploid). Variants are called by independently comparing each of the diploid assemblies to the reference.

The process is described in more detail in our Science paper. We also recommend that you read the Complete Genomics Service FAQ as background for this document.

Copy number variable (CNV) regions are called based on depth-of-coverage analysis. Sequence coverage is averaged and corrected for GC bias over a fixed window and normalized relative to a set of standard genomes. A hidden Markov model (HMM) is used to classify segments of the genome as having 0, 1, 2, 3 copies…up to a maximum value.

Structural variations (SVs) are detected by analyzing DNB mappings found during the standard assembly process described above and identifying clusters of DNBs in which each arm maps uniquely to the reference genome, but with an unexpected mate pair length or anomalous orientation. Local de novo assembly is applied to refine junction breakpoints and resolve the transition sequence. The process for CNV and SV detection is described in more detail in Complete Genomics Data File Formats.

In the summary file (summary-[ASM-ID].tsv), you will see a variety of metrics that may be helpful in understanding the quality of the delivered genome. For example:

• Gross Mapping Yield (Gb) — total base-pairs of sequence reads mapped to the reference genome (based on the initial mapping process only)
• Fully called genome fraction — percentage of reference genome with full (diploid) calls in the sequenced sample (following assembly)
• Fully called exome fraction — percentage of reference exome with full (diploid) calls in the sequenced sample (following assembly)
• Genome fraction where weightSumSequenceCoverage ≥n — Fraction of the reference genome bases where coverage is greater than or equal to n, with n being 5x, 10x, 20x, 30x and 40x.
• Exome fraction where weightSumSequenceCoverage ≥n — Fraction of the reference exome bases where coverage is greater than or equal to n, with n being 5x, 10x, 20x, 30x and 40x.

There are additional biological metrics which one would expect to be roughly consistent across genomes from individuals of the same ethnicity (even to genomes sequenced using other methods). These are also quite useful for quality control. They include:

• SNP total count (for genome and exome)
• SNP heterozygous/homozygous ratio (for genome and exome)
• SNP transitions/transversions ratio (for genome and exome)
• SNP novelty fraction (for genome and exome)

Please note that while the application of these and other metrics to normal diploid genomes is relatively clear, correctly interpreting these and similar calculations for a cancer or non-diploid genome can be more difficult.

In the REPORTS directory of our data delivery, you will find several files reporting various aspects of the sequence data that can be used to assess the quality of the delivered genome. For example:

• circos-[ASM-ID].html and circos-[ASM-ID].png (also: somaticCircos-[ASM-ID].html and somaticCircos-[ASM-ID].png): Shows a visual summary of small and large variation data for each genome. The image includes density of homozygous SNPs, density of heterozygous SNPs, gene symbols for impacted genes, and, when applicable, density of somatic variants.
• coverage-[ASM-ID].tsv: Reports number of bases in the reference genome covered (overlapped) by no reads, by one read, by two reads, etc. Two forms of coverage are computed and reported: uniquely mapping mated reads, and multiply mapping mated reads, appropriately weighted by a mapping confidence factor between 0 and 1 (“weight-sum” coverage). With this information, you can create a plot of genome-wide coverage distribution. For standard-coverage genomes, you would expect the mean coverage to be at least 40, and for high-coverage genomes the mean coverage would be at least 80.
• coverageCoding-[ASM-ID].tsv: Reports same information as coverage-[ASM-ID].tsv for only the coding regions of the reference genome.
• coverageByGcContent-[ASM-ID].tsv: Reports normalized coverage for cumulative GC base content percentile, allowing you to assess the level of GC bias across the genome.
• coverageByGcContentCoding-[ASM-ID].tsv: Reports normalized coverage for cumulative GC base content percentile, allowing you to assess the level of GC bias across the exome.

“Gross mapping yield” counts aligned bases within DNBs where at least one arm is mapped to the reference genome, excluding reads marked as overflow (large number of mappings to the reference genome indicative of highly repetitive sequence). “Both arms mapped yield” counts aligned bases within DNBs where both arms mapped to the reference genome on the correct strand and orientation and within the expected distance.

“Fully called” indicates that the assemblies of both diploid alleles meet the minimum required confidence thresholds, and thus both alleles are considered called. In this case, both alleles may be variant, or one may be reference and the other variant. If both are variant, they may be the same (homozygous) or different (heterozygous).

At a “partially called” or “half-called” site, only one allele meets the threshold to call the site confidently while the other does not. The Complete Genomics software reports this partial information for that locus (rather than no-calling the site entirely). Effectively, this is a statement that “we know this allele is present” but we can say little about what other allele is also present in a diploid region.

A “no-called” allele is one where we cannot determine the sequence of the sample at our minimum thresholds. See What exactly is a reference call? How is this different from a no-call?

The number of homozygous SNPs is calculated from the var-[ASM-ID].tsv file, and is equal to the sum of all diploid loci where the same SNP is present on both alleles.

The number of heterozygous SNPs is calculated from the var-[ASM-ID].tsv file, and is equal to the sum of SNPs present in the following types of loci:

• het-ref SNP: A single-base diploid locus where a SNP is present on one allele, and the other allele is reference.
• alt-alt SNP: A single-base diploid locus where each allele contains a different SNP.

The total number of SNPs is calculated from the var-[ASM-ID].tsv file, and includes SNPs present in all of the following types of loci:

• het-ref SNP: A single-base diploid locus where a SNP is present on one allele, and the other allele is reference.
• hom SNP: A single-base diploid locus where the same SNP is present on both alleles.
• alt-alt SNP: A single-base diploid locus where each allele contains a different SNP.
• hap SNP: A single-base haploid locus where a SNP is present on the single allele.
• other: All SNPs that occur in loci that do not fall into the above categories, including SNPs present in loci containing no-calls.

The exome is defined as the coding regions (CDS) of protein-coding genes, plus all of the untranslated genes, minus any transcripts (coding or otherwise) that are rejected by the annotation pipeline. A small percentage of transcripts in Build 36 and Build 37 are excluded from the annotation results due to the one or more of the following reasons:

• The transcript contains an unknown (“X”) amino acid.
• The start and/or stop codon positions are unknown.
• The transcript contains unspecified nucleotides.
• The transcript maps to an unknown location/chromosome.

To obtain the list of excluded transcripts, please contact support@completegenomics.com.

The number of exonic SNPs is calculated from the var-[ASM-ID].tsv file and includes SNPs present in all of the following categories:

• SYNONYMOUS: A SNP in the coding region of a transcript that has no affect on the protein sequence.
• NON-SYNONYMOUS: A SNP in the coding region of a transcript that alters protein sequence.
• DISRUPT: A SNP in the splice donor or splice acceptor sites of a transcript, predicted to disrupt splicing.
• NO-CHANGE: A SNP located outside the coding and splice donor/acceptor regions of a transcript, i.e., in the 3’ or 5’ untranslated regions (UTR3 or UTR5).
• UNKNOWN-TR: SNPs located in transcripts that are rejected by the annotation pipeline.

Each variation is only counted once per genome. If a variation has different functional consequencesin different transcripts, only the most severe functional effect is counted.

The following rules are used for counting variations:

• Variants in the pseudo-autosomal regions are only counted once.
• If a variant is contained within more than one transcript, we count only the most functionally deleterious mutation (we have a defined hierarchy).