BACKGROUND: This paper discusses the problem of automated annotation. It is a continuation of the previous work on the A4-algorithm (Adaptive algorithm of automated annotation) developed by Leontovich and others. RESULTS: A number of new statistics for the automated annotation of biological sequences is introduced. All these statistics are based on the likelihood ratio criterion. CONCLUSION: Some of the statistics yield a prediction quality that is significantly higher (up to 1.5 times higher) in comparison with the results obtained with the A4-procedure.

MOTIVATION: It is common knowledge that the avalanche of data arriving from the sequencing projects cannot be annotated either experimentally or manually by experts. The need for a reliable and convenient tool for automated sequence annotation is broadly recognized. RESULTS: Here, we describe the Adaptive Algorithm of Automated Annotation (A(4)) based on a statistical approach to this problem. The mathematical model relates a set of homologous sequences and descriptions of their functional properties, and calculates the probabilities of transferring a sequence description onto its homologue. The proposed model is adaptive, its parameters (distribution characteristics, transference probabilities, thresholds, etc.) are dynamic, i.e. are generated individually for the sequences and various functional properties (words of the description). The proposed technique significantly outperforms the widely used test for frequency threshold, which is a special case of our model realized for the simplest set of parameters. The prediction technique has been realized as a computer program and tested on a random sequence sampling from SWISS-PROT. AVAILABILITY: The automated annotation program based on the proposed algorithm is available through the Web browser at http://www.genebee.msu.su/services/annot/basic.html.