A joint NCI - DCRT evaluation of non-hierarchical cluster analysis has demonstrated that this technique is capable of providing a powerful tool for the investigation of multi-parameter flow cytometry data. Non-hierarchical cluster analysis routines, provided by Dr. Robert F. Murphy of Carnegie-Mellon University, were modified where necessary for our application and were applied to data from several list mode flow cytometry experiments which had been previously analyzed using conventional gated histogramming techniques. Direct application of the algorithms embodied in these routines has demonstrated the ability of cluster analysis to correctly partition complex cell populations and to identify small (less than 0.5% abundance) sub-populations, provided they are reasonably well separated in the 4-parameter data space. A graphical display package, also provided by Dr. Murphy, has been used to demonstrate a technique (N-plots) which enables the simultaneous visual evaluation of all recorded parameters for clusters or independent collections of individual cells. Based on the demonstrated capability of this analysis technique, a proposal was prepared for the development of a prototype advanced analysis facility for research oriented flow cytometry laboratories. This facility will integrate cluster analysis and conventional analysis techniques with a unified user interface and will maintain compatibility with existing standards in the field of flow cytometry.
