Algorithms are being developed and tested for aiding in numerical taxonomy of feature by strain matrices too large to be analyzed by existing programs. Both segmentation and heuristic approaches are being investigated. A program has been developed and is being tested to compare and evaluate methods and/or laboratories when characterizing the same set of strains. The usual statistical packages are not useful because of the predominantly binary (i.e., discontinuous) nature of the data. The algorithm allows comparison of tests or laboratories at the levels of the individual strain (with replicable determinations), species, genus, and overall set for determination of test method equivalences and/or inter-laboratory consistency. The program is being used to evaluate TB reference laboratories world-wide. A program for conversion of controlled vocabulary information in text records of the HDB into the highly compressed, table oriented MICRO-IS format has been implemented by the MSS for the HDB. The conversion process also has extensive spelling, syntax, and format error checking capabilities. A related project is the development of algorithms for format analysis and standardization of text images obtained by direct input of microbiological laboratory notebook information. Such facilities are required for computer database building of valuable archival paper records of phenotypic strain data. Computer graphic algorithms are being tested to aid microbiologists in visualizing individual similarities as well as hierarchical group memberships among strains.
