Re: The Digital "Disadvantage"?

From: Howard Shapiro (hms@shapirolab.com)
Date: Thu Jan 24 2002 - 21:29:23 EST


Jim (Jake) Jacobberger wrote-

>I'd like to weigh in on this one (from cytometry perspective and not
>on the sales or instrument specific issues). Several years ago I ran
>some cells through Leon Wheeless' slit scanning instrument that were
>stained with nucleolar antigens (p120 from Wade Bolton at Coulter) and
>DNA. Leon's group had for some time been deriving nuclear to cytoplasm
>ratios with acridine orange stained cells, and they have published
>extensively on this. In my experiment, the multiple nucleoli could
>easily be picked out of the traces, suggesting that significant
>sub-cellular localization was possible using antigen markers. Early
>work by Leon's group described an algorithmn for subtracting nuclear
>fluorescence from the entire trace (two dimensional analysis) that
>turns out to be mathmatically exact in three dimensions (unpublished),
>thus it appears that for some antigens (e.g., ~100% nuclear
>distribution), we could subtract ~exact non-specific staining on a
>cell by cell basis. Finally, sub-cellular translocalization is a major
>theme in cell signaling and cell cycle regulation and if possible
>should be a parameter (e.g., nuclear cyclin B1 should distinctly mark
>mitotic cells). Therefore, I have thought for some time that the best
>cytometer design is to use narrow beams, capture digital traces, and
>process/analzye the traces with cell biology in mind. However, I have
>a very defined end point (analysis of signaling molecules), and at the
>present time, other than the benefits described by Mario (dropping log
>amplifiers and easier compensation on multiparameter aquisition), with
>instruments that have large beams relative to cell size (DiVa) and
>without manufacturer software to capture files with traces (DiVa?? &
>Xcel), current instruments seem are a long way from what I want.

The digital pulse processing in the DiVa and Luminex instruments (and in
Rick Thomas's NPE system) was anticipated in Leon Wheeless's slit-scanning
flow cytometers decades ago, but the newer instruments (or at least the
DiVa and Luminex do not get a large enough number of samples per pulse to
permit much information to be derived from pulse shape.  The basic problem
is that both the DiVa and Luminex need to achieve a dynamic range in the
neighborhood of four decades, which requires at least a 14-bit
analog-to-digital converter (ADC).  The fastest sample rate available from
affordable 14-bit converters was, until recently, 10 MHz; the DiVa uses a
10 MHz converter, which means it can get 32 samples from a 3.2 usec pulse,
16 from a 1.6 usec pulse, 8 from an 800 nsec pulse, etc.  You'd probably
want at least 32 samples to do good shape analysis.  The options, not
mutually exclusive, are to flow at a lower velocity, giving longer pulses,
and to give up dynamic range in favor of more rapid sampling - modern
digital storage oscilloscopes use 8- and 10-bit ADC's that sample at
gigahertz rates.  David Galbraith and Jeff Rodriguez and their research
groups at the U. of Arizona collaborated on some work on pulse shape
analysis a few years back using fast ADC's; see Zilmer NA, Godavarti M,
Rodriguez JJ, Yopp TA, Lambert GM, Galbraith DW: Flow cytometric analysis
using digital signal processing.  Cytometry 20:102, 1995.

For what it's worth, Analog Devices now has a 14-bit, 105 MHz ADC.  This
could do pretty good pulse shape collection at conventional flow rates,
provided you could find a DSP chip to keep up with its data acquisition
rate.  The system design gets very complex; if a "killer application" -
perhaps Pap smears? - could be found, I'm sure somebody would find the time
and money to build the machine.

-Howard



This archive was generated by hypermail 2b29 : Wed Apr 03 2002 - 11:59:21 EST