Classification of ABC transporters using community detection

Abstract

The ATP-binding cassette (ABC) transporters are one of the major family of active transporters, present in the three domains of life. They are involved in the uptake and efflux of a wide variety of compounds, and thus play an essential role in the adaptation of organisms to their environment.We propose a new approach of ABC system classification exploiting the large amount of data available through systematic genome sequencing. Our main goal is to refine their functional prediction, i.e., which compound(s) is transported by a system. Our method relies on the identification of orthologous genes, i.e., genes that have diverged from a common ancestorafter an event of speciation and have retained the same function. These evolutionary links are inferred from sequence comparisons. A graph is constructed where vertices represent genes and edges orthology links. Our initial hypothesis is that highly connected isolated groups in our graph correspond to ABC genes involved in the transport of the same compound(s). However, we obtained much more complex graphs due to complex evolution scenarios. Thus, we have tested community detection algorithms to identify highly connected sets of orthologous genes in our graphs. The approach presented here enables us to study entire families of ABC transporters and to bring out communities that correlate well with subdivisions previously identified with another method. The communities obtained are then described in terms of function, i.e., specific compound(s) transported, using the functional annotations available in TCdb. © Springer-Verlag Berlin Heidelberg 2010.