Application of statistical and neural network techniques to biochemical data analysis

The methods of self-organising map (SOM) and principal components analysis (PCA) are investigated for analysing biochemical data generated in screening programmes to discover new pharmaceutically active compounds in microbial extracts. The organisms investigated, which belonged to the genus Streptomyces, had previously been classified by numerical taxonomy using a probabilistic identification matrix to one of 23 major taxonomic clusters based on 41 biological characters. The data related to extracts of organisms from 3 major clusters were available from 8 different pharmaceutical screens. When analysed by both techniques the data clustered into 3 groups which fitted well with their original taxonomic grouping. The results indicate the power of the techniques in analysing biological data and the predictive potential it may offer natural product screening programmes in the pharmaceutical industry.