Gunter has raised some very relevant issues with regard to data storage and management in flow cytometry. FCS 3.0 does an adequate, if cumbersome job, of providing a data interchange format. However, most instrumentation manufacturers developed software that would read FCS files that their instruments generated, and didn't put in much effort to covering other cases (one manufacturer, in particular, actively made it's software refuse to read FCS files from other sources, a stroke of "marketing brilliance"). This was facilitated by the design of FCS - the free form text header (keyword-values) contribute to the difficulty of developing a general library to parse these values into usable form. However, I don't believe that FCS per se is the cause of the difficulty Gunter expresses for his second criteria - interfacing with databases for patient or experiment management. A very important aspect has been consistently overlooked by commercial software - a method for users to easily fill in the necessary attributes for the FCS keywords. Much development has been spent on generating flashy graphics whose axis are "intelligently" labeled FL1 and FL2 (or P1 and P2). The combination of instrument configuration which would specify measurement parameter color, (FITC, TexRed, etc.), parameter scaling (log, lin), and other instrument conditions with sample attributes (source, cell type, stains, etc.) could currently all be stored in FCS keywords and read into various databases from them, satisfying mcuh of Gunter's second criteria. However without the development of appropriate software for users to generate this information, this won't happen. Moreover, I would think it appropriate for ISAC to generate some standards on what these attribute values should be. It would be very obnoxious to search your database for e.g. cd4 OR CD4 OR cd-4 OR CD-4 OR... As a "proof of principle" let me summerize what we have done here. Since 1983 we have provided a klunky (compared to GUIs) "protocol editor" which allowed users to generate a matrix of cell types and reagents. Our collection software associated each element of this matrix with one or more collected samples. This was combined, as indicated above, with instrument configuration parameters (from our standardization program). Data collected during one FACS run is accessed together as an "experiment" with an instrument, date, and user stamp. People could browse an archive of old experiments and bring up data with graphs whose axes were properly labeled and had reagent and cell type descriptions. Despite its unfriendliness, most Facility users found that the benefits outweighed the disadvantages and annotated their data. I think the lesson to be learned is that if the capacity to annotate exists and is relatively easy to use, most of the time it will be because the benefits are great. -Marty Bigos Stanford Shared FACS Facility
This archive was generated by hypermail 2b29 : Wed Apr 03 2002 - 11:49:56 EST