Assembling genome-wide transporter system annotations

Abstract

Over the years, several genome-scale metabolic models have been released with compartmentation information. Models such as iMH805/775 [1] (15 compartments) and iMM904 [2] (8 compartments) for Saccharomyces cerevisiae or iRS1563 [3] for Zea mays (6 compartments) include reactions performed in specific cellular organelles, such as mitochondria, chloroplasts (in photosynthetic organisms), lysosomes, cell nucleus or the Golgi apparatus, etc. Thus, cells have specific structures, the transport systems, to assist on the metabolites relocation. Cellular transport systems are described in databases such as TCDB (http://www.tcdb.org/) maintained by the Saier Lab Bioinformatics Group, the TransportDB (http://www.membranetransport.org/) or the YTPdb (http://homes.esat.kuleuven.be/~sbrohee/ytpdb/index.php/Main_Page) yeast transport protein database. TCDB proposed a classification system, analogous to the Enzyme Classification System (EC Number) [4] though including phylogenetic information, for transport proteins. The phylogenetic information included in the proposed nomenclature is very restrictive as it assigns a distinct TC Number to each transport protein identified in this database. Hence, in this work we propose a system to detect and classify potential transport proteins for a given genome. A TMHMM (http://www.cbs.dtu.dk/services/TMHMM/) [5] search was performed to identify potential transmembrane domains in the genome. Next, Smith-Waterman (SW) [6] alignments were performed on the transport candidate genes, against the TCDB database [7], to identify sequences encoding proteins with sequences similarities to the known transport systems. Instead of assigning a similar TC number to the transport system, the classification of the metabolites is proposed, according to the frequency of the metabolite in the SW search and the taxonomy resemblance of each hit organism to the case study organism, and not the transporter systems. The membrane where the transporter is located is predicted by the WoLF PSORT (http://wolfpsort.org/) [8] software.