be a false positive resulting from a portion of a read coming from a mRNA being aligned to
an intron adjacent to a splice junction. For end-distance bias, a t-test is performed in order
to determine whether the variants occur at a fixed distance from read ends, whereas for the
variant distance bias tests whether or not variant bases occur at random positions in the
aligned reads [19].
Indel and SNP Gap Filters These filters are designed to flag variants that are too close to
an indel or each other, respectively, as these may stem from alignment artifacts and
therefore be false positives. A value of 5 for these filters would mean that SNPs that are
five or fewer bp from an indel call would be discarded. The same would apply to clusters of
SNPs that are 5 or fewer bases apart from each other.
Review Question 3
Can you think of other possible filters that would need to be applied to the data postvariant calling to reduce the number of false positive calls?
These and other post-variant calling filters can be applied to a VCF file. Programs such
as bcftools [5] or GATK VariantFiltration [4] can be used for this purpose, and their
parameters can be tweaked to suit the researcher’s needs. Typically, after these are run, the
“FILTER” field in the VCF will be annotated as “PASS” if the variant has passed all
specified filters, or will have specified the filters that it has failed. Usually, the subsequent
analyses would be carried out only with those variants that passed all filters.
Now that we have seen the factors that researchers take into account when performing
an NGS study, we will next discuss the types of analyses they can follow to identify de
novo DNA variants or those that are likely to increase the risk of a disease.
10.5 De novo Genetic Variants: Population-Level Studies and
Analyses Using Pedigree Information
De novo mutations are crucial to the evolution of species and play an important role in
disease. De novo genetic variants are defined as those somatically arising during the
formation of gametes (oocytes, sperm) or that occur postzygotically. Only the mutations
present in germ cells can be transmitted to the next generation. Usually, when searching for
de novo variants in children (usually affected by developmental disorders), researchers
study trios, i.e. both parents and the child. The task is simplified because at 1.0 Â 1.10
À8 to
1.8 Â 1.10
À8 per nucleotide mutation rate, only a few de novo mutations are expected in the
germline of the child (a range of 44–88 according to [20]). If more than one variant fulfills
these criteria, bioinformatic methodologies such as examination of the extent of conservation throughout evolution, consequence prediction, and gene prioritization are then used to
pinpoint the most likely gene variants underlying the phenotype. Functional studies such as
10 Identification of Genetic Variants and de novo Mutations Based on NGS
133
an intron adjacent to a splice junction. For end-distance bias, a t-test is performed in order
to determine whether the variants occur at a fixed distance from read ends, whereas for the
variant distance bias tests whether or not variant bases occur at random positions in the
aligned reads [19].
Indel and SNP Gap Filters These filters are designed to flag variants that are too close to
an indel or each other, respectively, as these may stem from alignment artifacts and
therefore be false positives. A value of 5 for these filters would mean that SNPs that are
five or fewer bp from an indel call would be discarded. The same would apply to clusters of
SNPs that are 5 or fewer bases apart from each other.
Review Question 3
Can you think of other possible filters that would need to be applied to the data postvariant calling to reduce the number of false positive calls?
These and other post-variant calling filters can be applied to a VCF file. Programs such
as bcftools [5] or GATK VariantFiltration [4] can be used for this purpose, and their
parameters can be tweaked to suit the researcher’s needs. Typically, after these are run, the
“FILTER” field in the VCF will be annotated as “PASS” if the variant has passed all
specified filters, or will have specified the filters that it has failed. Usually, the subsequent
analyses would be carried out only with those variants that passed all filters.
Now that we have seen the factors that researchers take into account when performing
an NGS study, we will next discuss the types of analyses they can follow to identify de
novo DNA variants or those that are likely to increase the risk of a disease.
10.5 De novo Genetic Variants: Population-Level Studies and
Analyses Using Pedigree Information
De novo mutations are crucial to the evolution of species and play an important role in
disease. De novo genetic variants are defined as those somatically arising during the
formation of gametes (oocytes, sperm) or that occur postzygotically. Only the mutations
present in germ cells can be transmitted to the next generation. Usually, when searching for
de novo variants in children (usually affected by developmental disorders), researchers
study trios, i.e. both parents and the child. The task is simplified because at 1.0 Â 1.10
À8 to
1.8 Â 1.10
À8 per nucleotide mutation rate, only a few de novo mutations are expected in the
germline of the child (a range of 44–88 according to [20]). If more than one variant fulfills
these criteria, bioinformatic methodologies such as examination of the extent of conservation throughout evolution, consequence prediction, and gene prioritization are then used to
pinpoint the most likely gene variants underlying the phenotype. Functional studies such as
10 Identification of Genetic Variants and de novo Mutations Based on NGS
133