Gene analyses with VEGAS and PLINK were performed using the mean SNP statistic for VEGAS and both the mean SNP statistic (PLINK-avg) and the top SNP statistic (PLINK-top) for PLINK.

Analysis of summary SNP statistics

These SNP-wise models first analyse the individual SNPs in a gene and combine the resulting SNP p-values into a gene test-statistic, and can thus be used even when only the SNP p-values are available.

Although evaluation of the gene test-statistic does require an estimate of the LD between SNPs in the gene, estimates based on reference data with similar ancestry as the data the SNP p-values were computed from has been shown to yield accurate results [17,19].

SNPs were annotated to genes based on dbSNP version 135 SNP locations and NCBI 37.3 gene definitions.

For the main analyses only SNPs located between a gene’s transcription start and stop sites were annotated to that gene, yielding 13,172 protein-coding genes containing at least one SNP in the CD data.

This model first projects the SNP matriX for a gene onto its principal components (PC), pruning away PCs with very small eigenval-ues, and then uses those PCs as predictors for the phenotype in the linear regression model.

By default only 0.1% of the variance in the SNP data matriX is pruned away.

More traditional gene analysis models are also implemented, for comparison and to provide analysis of SNP summary statistics.

Efficient SNP to gene annotation and a batch mode for parallel processing are provided to simplify the overall analysis process.

We note that the modeling of sequencing errors is similar to the approach taken by some SNP calling methods, such as the one employed by GATK [27].

Reliable reference genomes are pivotal for finding genetic variations such as single nucleotide polymorphisms ( SNP ) and insertions/deletions (indels), which bolster downstream applications such as ge-nome-wide association studies, population genomics and comparative biology [4—7].

It is fundamentally different from copy-number detection algorithms, which are designed to look for departures from a normal situation of diploidy [24,25], and from SNP calling algorithms, which find variants based on the assumption of diploidy [26], or assume an user defined ploidy level [27,28].

Finally, the sensitivity of variant detection was high for all simulated situations, with false negative rates for SNP calling ranging from zero to 7.05%.

For example, a SNP with allele ratio 7:8 resulted in lower dosage accuracy than a SNP with allele ratio 13:2.

With higher coverage levels, both models had similar false negative rates of SNP discovery.

First, we define the probability of there being a SNP in any position as P( SNP ).

Overall, we called 134,464 variants within the contigs, with an average density of one SNP every 47 nucleotides.

We also performed variant calling with GATK [27] and obtained a density of one SNP every 60 bases.

We note that these SNP densities may be inflated due to homoeologue collapse and may not reflect exclusively allelic variation.

High concordance in SNP frequency between sequenced viral replicates from patients

To evaluate possible biases resulting from our RT-PCR procedure, we compared SNP frequencies in technical replicates, finding a high level of concordance.

In each case, the paired replicates showed SNP frequencies that were well correlated even when ignoring SNPs that occur with < 10% or >90% frequency in paired samples, R>0.95 for each pair (Fig 1).