Artificial neural networks have been combined with a rule based system to predict intron splice sites in the dicot plant Arabidopsis thaliana. A two step prediction scheme, where a global prediction of the coding potential regulates a cutoff level for a local prediction of splice sites, is refined by rules based on splice site confidence values, prediction scores, coding context, and distances between potential splice sites. In this approach, the prediction of splice sites mutually affect each other in a non-local manner. The combined approach drastically reduces the large amount of false positive splice sites normally haunting splice site prediction. An analysis of the errors made by the networks in the first step of the method revealed a previously unknown feature, a frequent T-tract prolongation containing cryptic acceptor sites in the 5' end of exons. The method presented here has been compared to three other approaches, GeneFinder, GeneMark, and Grail. Overall the method presented here is an order of magnitude better. We show that the new method is able to find a donor site in the coding sequence for the jelly fish Green Fluorescent Protein, exactly at the position that was experimentally observed in thaliana transformants. Predictions for alternatively spliced genes are also presented, together with examples of genes from other dicots, monocots, and algae. The method has been made available through electronic mail ( NetPlantGene@cbs.dtu.dk), or the WWW at http://www.cbs.dtu.dk/NetPlantGene.html
Keywords: Arabidopsis thaliana; splice site prediction; splice site pairing; plant biotechnology; neural networks; rule based systems.
Artificial neural networks have been applied to the prediction of splice site location in human pre-mRNA. A joint prediction scheme where prediction of transition regions between introns and exons regulates a cutoff level for splice site assignment was able to predict splice site locations with confidence levels far better than previously reported in the literature. The problem of predicting donor and acceptor sites in human genes is hampered by the presence of numerous amounts of false positives - in the paper the distribution of these false splice sites is examined and linked to a possible scenario for the splicing mechanism in vivo. When the presented method detects 95% of the true donor and acceptor sites it makes less than 0.1% false donor site assignments and less than 0.4% false acceptor site assignments. For the large data set used in this study this means that on the average there are one and a half false donor sites per true donor site and six false acceptor sites per true acceptor site. With the joint assignment method more than a fifth of the true donor sites and around one fourth of the true acceptor sites could be detected without accompaniment of any false positive predictions. Highly confident splice sites could not be isolated with a widely used weight matrix method or by separate splice site networks. A complementary relation between the confidence levels of the coding/non-coding and the separate splice site networks was observed, with many weak splice sites having sharp transitions in the coding/non-coding signal and many stronger splice sites having more ill-defined transitions between coding and non-coding.