Speakers
Description
Abstract
We analyze plant genome data with the aim of understanding genome evolution and linking genome information to plant traits, thereby also supporting breeding of better performing and more resilient crops.
Bioinformatics integrates sequence information, mainly from DNA sequencing, with high-throughput data for example from genomics and phenotyping. The data life cycle in bioinformatics relies on primary analysis of data, leading e.g. to sequences or phenotype measurements, but then several layers of analysis transforming this primary data to biologically meaningful secondary data, e.g. genes, proteins, traits. This requires bioinformatics prediction methods that are heterogeneous and evolving. In order to deduce scientific insights, data integration is required.
Data management faces several challenges in the bioinformatics data life cycle:
1) Lack of ground truth: In the post-genomic era, where DNA is sequenced faster than its function is described, functional annotations today are mostly predicted and only a fraction is experimentally verified in the lab. This limits the assessment of sensitivity and accuracy of the employed methods. 2) Reproducibility depends on the documentation of details that are easy to lose track of: exact tool versions, parameter choices, reference genome builds. 3) Missing metadata limits knowledge inference from data integration. Without consistent information on origins, samples and treatments, data integration might, in the worst case, even lead to wrong biological conclusions.
Due to these challenges, the re-use of public data is tedious: A lot of quality control and re-analysis is necessary to control biases in order to draw meaningful insights from the data.
While none of the issues presented are new, with the speed the field is developing, solutions need to be adapted and improved on constantly. Here we aim to give an overview of the current problems we face and the solutions we apply, as well as larger scale solutions we propose for the community.