Building a livestock genetic and genomic information knowledgebase through integrative developments of Animal QTLdb and CorrDB

Abstract

Successful development of biological databases requires accommodation of the burgeoning amounts of data from high-throughput genomics pipelines. As the volume of curated data in Animal QTLdb (https://www.animalgenome.org/QTLdb) increases exponentially, the resulting challenges must be met with rapid infrastructure development to effectively accommodate abundant data curation and make metadata analysis more powerful. The development of Animal QTLdb and CorrDB for the past 15 years has provided valuable tools for researchers to utilize a wealth of phenotype/genotype data to study the genetic architecture of livestock traits. We have focused our efforts on data curation, improved data quality maintenance, new tool developments, and database co-developments, in order to provide convenient platforms for users to query and analyze data. The database currently has 158 499 QTL/associations, 10 482 correlations and 1977 heritability data as a result of an average 32% data increase per year. In addition, we have made >14 functional improvements or new tool implementations since our last report. Our ultimate goals of database development are to provide infrastructure for data collection, curation, and annotation, and more importantly, to support innovated data structure for new types of data mining, data reanalysis, and networked genetic analysis that lead to the generation of new knowledge.

Figure 1.

Growth of curated data in the Animal QTLdb by year and species, based on the total from three data releases per year on the Animal QTLdb website. Note that all data are log transformed so that the bar graphs can fit into a reasonable window size. (A) Animal QTLdb data growth by species. (B) Animal QTLdb data growth by data publication year and data type (note the different scale for the association data plot on the right axis). Please note that this is a reflection of data growth in the database, not a measure of all data in the public domain. There are data from recent years still in the curation pipeline that are not counted here.

Figure 2.

Example output from the QTLdb data enrichment analysis tool. The analysis was performed on 3827 ‘milk yield’ QTL/associations found in cattle (these milk traits represent a collection of seven related traits measured/estimated with different methods, each describing certain aspects of the ‘milk yield’). A Chi-squared analysis was performed on a 7 × 30 contingency table. The results show p-values for each chromosome, along with false discovery rate (FDR) values estimated using the Benjamini–Hochberg procedure. The transformed values of Chi-squares are plotted using horizontal bars to indicate locations where larger numbers of QTL/associations are found.

Figure 3.

A screenshot of the trait mapping tool used in the Animal QTLdb and CorrDB. The tool accommodates comparative views of the three ontologies against the livestock trait set, so that the best match can be chosen while providing feedback for ontology developments.

Figure 4.

Data links between the Animal QTLdb and CorrDB are achieved based on their mutual trait mapping to VT/LPT/CMO ontologies. (A) A QTL/association data view showing links to CorrDB where they exist (highlighted in light green). (B) A CorrDB view of correlations showing traits with existing QTL/association data.

Figure 5.

A diagram showing how the ‘modifiers’ are used to annotate traits in the environment of ontology management of hierarchies. Use of the modifiers effectively allows multidimensional attributes to be appended to a trait.

Figure 6.

Screenshots of the trait editor tools showing how trait modifiers are managed with controlled vocabulary and context (A), and the formulation of a trait variant by adding modifiers to more clearly define how a trait may be evaluated in a given context (B). (A) A QTLdb editor window showing how ‘modifier’ attributes are managed. (B) A QTLdb editor window showing how a trait variant with modifiers can be created.

Figure 7.

Gene-centric (A) and trait-centric (B) views of animal QTL/association data. Note that the long QTL/association list is hidden upon first loading of the page. This allows users a quick view of the gene or trait list before expanding the details on a particular item for closer examination or data download. (A) Gene-centric view of lists of QTL/association data. (B) Trait-centric view of lists of QTL/association data.