Identification of a new motif on nucleic acid sequence data using Kohonen's self-organizing map

Here we present a performance test of a Kohonen features mapapplied to the fast extraction of uncommon sequences from thecoding region of the human insulin receptor gene. We used anetwork with 30 neurons and with a variable input window. Theprogram was aimed at detecting unique or uncommon DNA regionspresent in crude sequence data and was able to automaticallydetect the signal peptide coding regions of a set of human insulinreceptor gene data. The testing of this program with HSIRPRcDNA release (EMBL data bank) indicated the presence of uniquefeatures in the signal peptide coding region. On the basis ofour results this program can automatically detect ‘singularity’from crude sequencing data and it does not require knowledgeof the features to be found. Received on August 27, 1990; accepted on March 14, 1991