High speed sorting

Brian Hall bhall at csn.org
Mon Apr 3 14:59:41 EST 1995

Mike Salmon, from Birmingham UK, responded to Mark Cameron's request for
information on high speed sorters:
You may well have quotes, but I would be surprised if they have the
machines to sort at 20,000 cells per second.  Whatever different companies
claim as a sorting rate most people I know who use machines from any
of these companies work nearer 1000 cells per second if they need reasonably
clean populations.
Have fun
Mike Salmon"

This network should not be a medium for sales propoganda, but Mike's
comments demand a reply from Cytomation.

Mike's observation is generally true.  But he has unfortunately lumped
Cytomation's MoFlo in with the machines of which he has experience.  With
respect, Mike, you do not know anyone with a machine from Cytomation, yet.
When you do you will know someone with a machine easily capable of
sorting at well over 20,000 per second, with high purity AND high yield,
AND with more than one beam.

MoFlo was engineered from the beginning to operate at pressures up to 100
psi, to form droplets at 100kHz or more, to sort efficiently with only one
drop per cell, to cope with trigger rates far higher than 20,000 per
second without errors and to cope simultaneously with data from more than
one spatially separated beam without error or data loss.  All of these
features are essential to real high speed sorting.

the number most often missing or under-emphasized in claims of high speed
is yield.  We consistently achieve yields around 80 percent at 20,000 per
second on MoFlo, with rate populations.  If you run at 20,000 per second
and 90 percent of the cells are too close together to be separated by the
instrument (either electronically or into droplets), you may as well be
running at 2,000 per second.  That is the reason for Mike's observation
that few people can sort faster than 1,000 per second if they want clean
(pure) sorts.  And, on all machines other than MoFlo, the yield drops even
more dramatically when more than one beam is used.

I have heard two arguments in defense of low yield at high trigger rates
when coincidence aborts are enabled.  One is that recovery overrides
purity in many high speed applications and that coincidence aborts are
therefore often unnecessary.  However, even in "enrichment" more, where
all (rare) target cells are sorted even if unwanted cells come along, a
well-designed instrument results in a useful improvement in purity
(about 20 percentage points for a 100kHz droplet formation machine over a
30kHz machine at 20,000 triggers per second).

I also have heard proposed the rather strange notion that rare cell
sorting (the most common reason for seeking high speed sorts) is "low
speed sorting" because the number of sorts per second is low even when the
trigger rate is very high (for example, .01% of 20,000 per second is just
2 sorts per second).  This argument goes that the true test of a sorter
is sorting fractions that are 50 percent or more of the total sample,
where the number of "yes" sort decisions per second is high.

The fact is that the instrument has to process and decide on every cell
and is no more stressed if it deflects every cell it sees than if it
deflects none at all.  In fact, rare cell sorting demands a better
instrument, for the following reason.  When the target population is rare,
almost every concident event must be aborted and the yield falls away
rapidly if the instrument is not designed to minimize coincidences.

Brian Hall
Cytomation, Inc.
800 822-9902 or
(303) 226-2200
Fax: (303) 226-0107

More information about the Cytometry mailing list