four criteria: If the variant had been seen in other cardiomyopathy patients, if it was absent
from 200 alleles from controls, if it was conserved among species and isoforms and if it cosegregated with the disease in affected families. However, the chance of all these criteria
being fulfilled by chance alone if any other genes had been considered is high—as a study
subsequently found by assessing a larger gene panel and calculating the expected number
of variants in the gene [34]. Additionally, both positive and negative evidence for the
hypothesis should be carefully evaluated, for example, in the same cardiomyopathy study
some of the “potentially causal” variants predicted by bioinformatics algorithms did not cosegregate with the phenotype [34]. The increasing availability of sequencing data in large
cohorts such as gnomAD should help establishing causality as more accurate allele
frequencies are reported per population [15]. This is an important point—allele frequencies
should be matched by ancestry as closely as possible, as it is known that they can vary
greatly among different populations [25].
Another important set of criteria, highlighted by MacArthur et al [32], argues that when
analyzing potentially monogenic diseases, genes that have previously been confidently
linked to similar phenotypes should be analyzed as the first potential candidates before
proceeding to explore novel genes, and that if a researcher does proceed to analyzing
further genes, then multiple independent carrier individuals must present with similar
clinical phenotypes. Additionally, it is desirable that the distribution of variants in a suitable
control population is examined, for example, if a researcher has identified a novel stopgained variant in a candidate gene, how many other stop-gained variants are found in
population-level variation catalogues?
Finally, statistical evidence and multiple computational approaches may strongly suggest that a variant is disease-causing. However, whenever possible, researchers should
perform functional studies that indicate this is the case, whether by using tissue derived
from patients themselves, cell lines, or model organisms. The comprehensive view
provided by statistical, computational, and functional studies then may be enough for a
researcher to report a potential causal variant. In doing so, it is recommended that all
available evidence is detailed, clear and uncertain associations are reported and that all
genetic data is released whenever possible [32].
10.6.3 Variant Filtering and Visualization Programs
Finally, visual representation of genomic data can be highly useful for the interpretation of
results [28]. Visualization tools can help users browse mapped experimental data along
with annotations, visualize structural variants, and compare sequences. These programs can
be available as stand-alone tools or as web applications, and vary in the amount of
bioinformatics knowledge necessary to operate them. Here we will review some of the
most popular and that we consider useful, but there are many others suited for different
purposes and with a range of functionalities.
136
P. Basurto-Lozada et al.
from 200 alleles from controls, if it was conserved among species and isoforms and if it cosegregated with the disease in affected families. However, the chance of all these criteria
being fulfilled by chance alone if any other genes had been considered is high—as a study
subsequently found by assessing a larger gene panel and calculating the expected number
of variants in the gene [34]. Additionally, both positive and negative evidence for the
hypothesis should be carefully evaluated, for example, in the same cardiomyopathy study
some of the “potentially causal” variants predicted by bioinformatics algorithms did not cosegregate with the phenotype [34]. The increasing availability of sequencing data in large
cohorts such as gnomAD should help establishing causality as more accurate allele
frequencies are reported per population [15]. This is an important point—allele frequencies
should be matched by ancestry as closely as possible, as it is known that they can vary
greatly among different populations [25].
Another important set of criteria, highlighted by MacArthur et al [32], argues that when
analyzing potentially monogenic diseases, genes that have previously been confidently
linked to similar phenotypes should be analyzed as the first potential candidates before
proceeding to explore novel genes, and that if a researcher does proceed to analyzing
further genes, then multiple independent carrier individuals must present with similar
clinical phenotypes. Additionally, it is desirable that the distribution of variants in a suitable
control population is examined, for example, if a researcher has identified a novel stopgained variant in a candidate gene, how many other stop-gained variants are found in
population-level variation catalogues?
Finally, statistical evidence and multiple computational approaches may strongly suggest that a variant is disease-causing. However, whenever possible, researchers should
perform functional studies that indicate this is the case, whether by using tissue derived
from patients themselves, cell lines, or model organisms. The comprehensive view
provided by statistical, computational, and functional studies then may be enough for a
researcher to report a potential causal variant. In doing so, it is recommended that all
available evidence is detailed, clear and uncertain associations are reported and that all
genetic data is released whenever possible [32].
10.6.3 Variant Filtering and Visualization Programs
Finally, visual representation of genomic data can be highly useful for the interpretation of
results [28]. Visualization tools can help users browse mapped experimental data along
with annotations, visualize structural variants, and compare sequences. These programs can
be available as stand-alone tools or as web applications, and vary in the amount of
bioinformatics knowledge necessary to operate them. Here we will review some of the
most popular and that we consider useful, but there are many others suited for different
purposes and with a range of functionalities.
136
P. Basurto-Lozada et al.