Representativeness of variation benchmark datasets.

BACKGROUND: Benchmark datasets are essential for both method development and performance assessment. These datasets have numerous requirements, representativeness being one. In the case of variant tolerance/pathogenicity prediction, representativeness means that the dataset covers the space of variations and their effects. RESULTS: We performed the first analysis of the representativeness of variation benchmark datasets. We used statistical approaches to investigate how proteins in the benchmark datasets were representative for the entire human protein universe. We investigated the distributions of variants in chromosomes, protein structures, CATH domains and classes, Pfam protein families, Enzyme Commission (EC) classifications and Gene Ontology annotations in 24 datasets that have been used for training and testing variant tolerance prediction methods. All the datasets were available in VariBench or VariSNP databases. We tested also whether the pathogenic variant datasets contained neutral variants defined as those that have high minor allele frequency in the ExAC database. The distributions of variants over the chromosomes and proteins varied greatly between the datasets. CONCLUSIONS: None of the datasets was found to be well representative. Many of the tested datasets had quite good coverage of the different protein characteristics. Dataset size correlates to representativeness but only weakly to the performance of methods trained on them. The results imply that dataset representativeness is an important factor and should be taken into account in predictor development and testing.

Large differences in proportions of harmful and benign amino acid substitutions between proteins and diseases.

Genes and proteins are known to have differences in their sensitivity to alterations. Despite numerous sequencing studies, proportions of harmful and harmless substitutions are not known for proteins and groups of proteins. To address this question, we predicted the outcome for all possible single amino acid substitutions (AASs) in nine representative protein groups by using the PON-P2 method. The effects on 996 proteins were studied and vast differences were noticed. Proteins in the cancer group harbor the largest proportion of harmful variants (42.1%), whereas the non-disease group of proteins not known to have a disease association and not involved in the housekeeping functions had the lowest number of harmful variants (4.2%). Differences in the proportions of the harmful and benign variants are wide within each group, but they still show clear differences between the groups. Frequently appearing protein domains show a wide spectrum of variant frequencies, whereas no major protein structural class-specific differences were noticed. AAS types in the original and variant residues showed distinctive patterns, which are shared by all the protein groups. The observations are relevant for understanding genetic bases of diseases, variation interpretation, and for the development of methods for that purpose.

VariOtator, a Software Tool for Variation Annotation with the Variation Ontology.

The Variation Ontology (VariO) is used for describing and annotating types, effects, consequences, and mechanisms of variations. To facilitate easy and consistent annotations, the online application VariOtator was developed. For variation type annotations, VariOtator is fully automated, accepting variant descriptions in Human Genome Variation Society (HGVS) format, and generating VariO terms, either with or without full lineage, that is, all parent terms. When a coding DNA variant description with a reference sequence is provided, VariOtator checks the description first with Mutalyzer and then generates the predicted RNA and protein descriptions with their respective VariO annotations. For the other sublevels, function, structure, and property, annotations cannot be automated, and VariOtator generates annotation based on provided details. For VariO terms relating to structure and property, one can use attribute terms as modifiers and evidence code terms for annotating experimental evidence. There is an online batch version, and stand-alone batch versions to be used with a Leiden Open Variation Database (LOVD) download file. A SOAP Web service allows client programs to access VariOtator programmatically. Thus, systematic variation effect and type annotations can be efficiently generated to allow easy use and integration of variations and their consequences.

Human Variome Project Quality Assessment Criteria for Variation Databases.

Numerous databases containing information about DNA, RNA, and protein variations are available. Gene-specific variant databases (locus-specific variation databases, LSDBs) are typically curated and maintained for single genes or groups of genes for a certain disease(s). These databases are widely considered as the most reliable information source for a particular gene/protein/disease, but it should also be made clear they may have widely varying contents, infrastructure, and quality. Quality is very important to evaluate because these databases may affect health decision-making, research, and clinical practice. The Human Variome Project (HVP) established a Working Group for Variant Database Quality Assessment. The basic principle was to develop a simple system that nevertheless provides a good overview of the quality of a database. The HVP quality evaluation criteria that resulted are divided into four main components: data quality, technical quality, accessibility, and timeliness. This report elaborates on the developed quality criteria and how implementation of the quality scheme can be achieved. Examples are provided for the current status of the quality items in two different databases, BTKbase, an LSDB, and ClinVar, a central archive of submissions about variants and their clinical significance.

VariSNP, a benchmark database for variations from dbSNP.

For development and evaluation of methods for predicting the effects of variations, benchmark datasets are needed. Some previously developed datasets are available for this purpose, but newer and larger benchmark sets for benign variants have largely been missing. VariSNP datasets are selected from dbSNP. These subsets were filtered against disease-related variants in the ClinVar, UniProtKB/Swiss-Prot, and PhenCode databases, to identify neutral or nonpathogenic cases. All variant descriptions include mapping to reference sequences on chromosomal, genomic, coding DNA, and protein levels. The datasets will be updated with automated scripts on a regular basis and are freely available at http://structure.bmc.lu.se/VariSNP.

Update of PAX2 mutations in renal coloboma syndrome and establishment of a locus-specific database.

Renal coloboma syndrome, also known as papillorenal syndrome is an autosomal-dominant disorder characterized by ocular and renal malformations. Mutations in the paired-box gene, PAX2, have been identified in approximately half of individuals with classic findings of renal hypoplasia/dysplasia and abnormalities of the optic nerve. Prior to 2011, there was no actively maintained locus-specific database (LSDB) cataloguing the extent of genetic variation in the PAX2 gene and phenotypic variation in individuals with renal coloboma syndrome. Review of published cases and the collective diagnostic experience of three laboratories in the United States, France, and New Zealand identified 55 unique mutations in 173 individuals from 86 families. The three clinical laboratories participating in this collaboration contributed 28 novel variations in 68 individuals in 33 families, which represent a 50% increase in the number of variations, patients, and families published in the medical literature. An LSDB was created using the Leiden Open Variation Database platform: www.lovd.nl/PAX2. The most common findings reported in this series were abnormal renal structure or function (92% of individuals), ophthalmological abnormalities (77% of individuals), and hearing loss (7% of individuals). Additional clinical findings and genetic counseling implications are discussed.

LOVD v.2.0: the next generation in gene variant databases.

Locus-Specific DataBases (LSDBs) store information on gene sequence variation associated with human phenotypes and are frequently used as a reference by researchers and clinicians. We developed the Leiden Open-source Variation Database (LOVD) as a platform-independent Web-based LSDB-in-a-Box package. LOVD was designed to be easy to set up and maintain and follows the Human Genome Variation Society (HGVS) recommendations. Here we describe LOVD v.2.0, which adds enhanced flexibility and functionality and has the capacity to store sequence variants in multiple genes per patient. To reduce redundancy, patient and sequence variant data are stored in separate tables. Tables are linked to generate connections between sequence variant data for each gene and every patient. The dynamic structure allows database managers to add custom columns. The database structure supports fast queries and allows storage of sequence variants from high-throughput sequence analysis, as demonstrated by the X-chromosomal Mental Retardation LOVD installation. LOVD contains measures to ensure database security from unauthorized access. Currently, the LOVD Website (http://www.LOVD.nl/) lists 71 public LOVD installations hosting 3,294 gene variant databases with 199,000 variants in 84,000 patients. To promote LSDB standardization and thereby database interoperability, we offer free server space and help to establish an LSDB on our Leiden server.

BTK gatekeeper residue variation combined with cysteine 481 substitution causes super-resistance to irreversible inhibitors acalabrutinib, ibrutinib and zanubrutinib.

Irreversible inhibitors of Bruton tyrosine kinase (BTK), pioneered by ibrutinib, have become breakthrough drugs in the treatment of leukemias and lymphomas. Resistance variants (mutations) occur, but in contrast to those identified for many other tyrosine kinase inhibitors, they affect less frequently the "gatekeeper" residue in the catalytic domain. In this study we carried out variation scanning by creating 11 substitutions at the gatekeeper amino acid, threonine 474 (T474). These variants were subsequently combined with replacement of the cysteine 481 residue to which irreversible inhibitors, such as ibrutinib, acalabrutinib and zanubrutinib, bind. We found that certain double mutants, such as threonine 474 to isoleucine (T474I) or methionine (T474M) combined with catalytically active cysteine 481 to serine (C481S), are insensitive to ≥16-fold the pharmacological serum concentration, and therefore defined as super-resistant to irreversible inhibitors. Conversely, reversible inhibitors showed a variable pattern, from resistance to no resistance, collectively demonstrating the structural constraints for different classes of inhibitors, which may affect their clinical application.