Non-invasive prenatal testing as a valuable source of population specific allelic frequencies

Abstract

Low-coverage massively parallel genome sequencing for non-invasive prenatal testing (NIPT) of common aneuploidies is one of the most rapidly adopted and relatively low-cost DNA tests. Since aggregation of reads from a large number of samples allows overcoming the problems of extremely low coverage of individual samples, we describe the possible re-use of the data generated during NIPT testing for genome scale population specific frequency determination of small DNA variants, requiring no additional costs except of those for the NIPT test itself. We applied our method to a data set comprising of 1,548 original NIPT test results and evaluated the findings on different levels, from in silico population frequency comparisons up to wet lab validation analyses using a gold-standard method. The revealed high reliability of variant calling and allelic frequency determinations suggest that these NIPT data could serve as valuable alternatives to large scale population studies even for smaller countries around the world. Langmead B, Salzberg SL. 2012. Fast gapped-read alignment with Bowtie 2. Nature methods frequencies of variants in our sample set and six different ExAC populations. Results are shown separately for (a) SNVs based on 68,326 variants; and (b) indels based on 2,909. In both cases, only those variants were used, which were simultaneously identified in each data subset with MAF higher than 5%. AFR = African/African American, AMR = American (Latino), EAS = East Asian, FIN = Finnish, NFE = Non-Finnish European, SAS = South Asian, SVK = Slovak. 2 7 SupplementaryTable 1 (uploaded as a separate Excel sheet of the Supplementary Table file):List of variants identified in the region of the CLCN1 (chloride voltage-gated channel 1, OMIM * 118425) gene in ExAC populations, dbSNP and in our data set. Only 17 of them had ExAC frequencies above 5% (highlighted in yellow). All but five were identified in our data set too with very similar calculated population frequencies. The exceptions, rs34904831, rs191902231, rs182668076, rs2280663 and rs73726622, were found to have ExAC frequencies +/-5% (depending on population). Originally three of these variants were identified in our data set too, although they were filtered out due slightly lower than 5% frequency (Suppl.Tab.1), further suggesting the feasibility of lowering our frequency restrictions. Our data contained also 89 variants missing from ExAC (highlighted in red), but found to have each of them deep intronic positions falling outside ExAC´s BED file. AFR = African/African American, AMR = American, EAS = East Asian, FIN = Finnish, NFE = Non-Finnish European, SAS = South Asian, SVK = Slovak. Supplementary Table 2 (uploaded as a separate Excel sheet of the Supplementary Table file): Verification of 87 positions in five polymorphic genomic loci of 58 randomly selected samples of our sample set from which genomic DNA was available to validation purposes. Reads from positions covered by multiple reads in individual samples (such in case of sample ID4730, where both alleles we...

Genetic profiling of 2,683 Vietnamese genomes from non-invasive prenatal testing data

“…Although the sequencing depth per sample was low, there might be more than one read from the same sample overlapping at a genome position. This problem may affect the estimation of allele frequency because the estimation is based on the assumption that a sample may contribute only 0 or 1 allele (read) at any given genome position [12,13]. For instance, we found that the average percentage of genome positions with depth 2x (i.e.…”

“…At any genome position, there were on average 47 out of 2,683 samples that each contributed two reads (Supplementary Figure S3). To address this problem, we followed a filtering strategy from previous studies [12,13] to keep only one read if there were overlapping reads in a sample. Thus, for every genome position, each sample could only contribute up to one read, and when the samples were aggregated, all reads at any position were obtained from different samples.…”

“…We aggregated 2,683 samples into one and used Mutect2 from GATK [26,27] for variant calling and allele frequency estimation. In addition to its main function of somatic calling, Mutect2 can also be used on data that represents a pool of individuals, such as our NIPT dataset, to call multiple variants at a genome site [12,13,28]. The called variants were further checked against strand bias, weak evidence, or contamination using FilterMutectCalls.…”

“…An alternative approach has been proposed recently to re-use the low-coverage genome sequencing data from non-invasive prenatal testing (NIPT) for large-scale population genetics studies [12,13]. NIPT is a method that sequences cell-free DNA from maternal plasma at an ultra-low depth of 0.1-0.2x to detect fetal aneuploidy [14].…”

“…In this paper, we presented the first study of Vietnamese genetic variations from non-invasive prenatal testing data, and to the best of our knowledge, the third of such kinds in the world [12,13]. We analyzed the NIPT data of 2,683 pregnant Vietnamese women to identify genetic variants and their allele frequency distribution in the Vietnamese population.…”