| Flow cytometry (FCM) is a powerful technique to measure multiple characteristics of individual particles, usually cells, in order to facilitate medical diagnosis and treatment, vaccine development, and even in cross-matching organs for transplantation. Nowadays the availability of FCM analysis autogating algorithms has made the analysis of high throughput and high dimensional FCM data available without subjective and labor intensive manual work. However, cross-sample comparison, the next step of the analysis chain, remains a big challenge, since the two prevalent methods of cross sample comparison available, the centroid clustering algorithm and FLAME algorithm, have their inherent disadvantages. Therefore, we propose a solution based on Friedman-Rafsky Test (F-R test) to satisfy the growing need for the development of accurate methods.;We conducted a simulation study to access the performance of the F-R test and concluded that the F-R test has outstanding performance in the identification of matching SPs across samples where we have rare SPs, abundant SPs, SPs similar in location, or a different number of SPs in two samples. The F-R test can accurately match SPs when one or more of the samples have been over/under partitioned, which indicates that it can be used to diagnose over/under partitioning in autogating algorithms. Furthermore, the F-R test is robust to moderate shifts in the mean of a subpopulation (as might occur when a cytometer is not properly calibrated) except for some rare and dense SPs. In the case of absolute position changes of all the SPs, the performance of F-R test can be improved significantly by normalizing the data before applying the F-R test. We illustrated the application of F-R test through a real dataset and compared the results with those of centroid clustering algorithm and FLAME algorithm. The F-R test outperformed centroid clustering and FLAME algorithms with nearly perfect matching results.;In conclusion, we believe that this F-R test based method could facilitate the application of FCM to disease diagnostic and drug development with more accurate cross sample comparison results. |
