Database Construction And Precursor Prediction For Neuropeptide

Neuropeptides(NPs) play critical roles in synaptic signaling in all nervous systems. NPs governing broad brain functions such as analgesia, food intake, metabolism, reproduction, social behaviors, learning and memory, and act as a number of roles such as hormone, modulators, transmitter of neuro and cytokines. NPs share the common characteristic that is produced from a longer NP precursor(NPP). Generally, one NPP includes one signal peptide, one or more neuropeptide(s) and other sequences. The study focus on neuropeptide is always a hotpot field of bioinformatics. But current neuropeptide databases lack authority evaluation of innovation and accuracy. Furthermore, there are some questions about database, for example, the delayed update leads to the deficiency of data, and the data is redundant and repeated. So the relative database is far insufficient to support follow-up studies. Meanwhile, the basic study should be supported powerfully by many neuropeptide predictive tools which have higher accuracy. So, this paper does the research from the following two aspects:First, based on the present studies we have developed a comprehensive resource of NPs with data is richer and information is more complete, named NeuropepDB. With expanding existing NP databases and extracting the NPs from the published papers, Neuropep DB holds 2545 non-redundant neuropeptide precursors and 6275 neuropeptide entries originating from 442 organisms belonging to 65 neuropeptide families. 4936 of 6275 NPs are resulting from complex and variable posttranslational processing of NPP, and the rest of 1341 NPs are not included in SWISS-PROT collecting from 342 papers.Second, developed NeuroPP(Neuropeptide Precursor Predictor) based on support vector machine to forecast the NPPs. NeuroPP integrated two predictors, one was derived on dipeptide features and another was derived on tripeptide features. The algorithm makes optimization about subspace by using the analysis of variance method to choose higher contribution features and ignoring lower contribution features from dipeptides and tripeptides, using the optimal attributes subset of features to develop the prediction model. Evaluated with five-fold cross-validation and independent datasets, NeuroPP shows good performance that achieves an accuracy of 82.52% with a ROC of 0.90, indicating that it is preforms splendidly in recognition NPPs and is a helpful complement to available tools.For the convenience exchange and query dates, the NeuropepDB is freely available at i.uestc.edu.cn/neuropeptide, the NeuroPP as a tool of the NeuropepDB has been run successfully at i.uestc.edu.cn/neuropeptide/cgi-bin/NeuroPP.pl.