3700 Data Collection v2 extractor can invisibly delete sequence.

Phillip San Miguel pmiguel at purdue.edu
Sun May 11 11:42:49 EST 2003

    Just a "heads-up" for those using Data Collection v. 2 on the 3700
for sequencing. While a great improvement over v 1.1 in many ways, it
does appear to have one characteristic that I find extremely
undesirable--even frightening. Many long-time users of the 3700 will
know that under certain circumstances (laser or camera problems) there
will be periods of time were signal drops to nothing across the array
then comes back later on. The result in the array view pane of Data
Collection is a band of black across normal sequence. I can only vaguely
remember the previous time this occurred (years ago)--but I think these
areas with no data were clearly visible in the sequence produced--both
text and chromatograms.
    This week I think my laser has been cycling off every 6 or 7 blocks.
(I noticed yesterday that it was drawing 10.6 mA at 40 mW power--it
should draw less than 9 mA.) While this is merely irritating, I would
expect that it would be apparent after the first run this was happening
(20-30 bases of N's every 150-200 base of sequence, for example.) But
with version 2 of data collection I see no such thing. The regions with
no sequence are *spliced out*. Further curve smoothing during processing
makes the gaps in the sequence undetectable when one looks at the
processed sequence. Even looking at the raw sequence chromatograms using
Sequence Analysis the spliced sequence gaps are hard to see--mainly one
might note periodic drops in the base line at the splice sites.
    To spell out why this is bad, the result is that sequences generated
will have 20-30 base deletions every 150-200 bases (in my case). You
won't see anything odd by looking at the chromatograms--only by paging
through a run in Data Collection will one clearly see blank areas.
    My advice: if you automatically use phred on your runs be on the
look-out for runs where most of the sequences are shorter than normal
but the number of high quality bases is not drastically decreased. My
total read lengths are generally around 900 bases with the number of
high quality bases being between 600 and 800 (on a good run). This week
I was seeing 750 bases total with the number of high quality bases being
400-600 in most cases. (I suggest looking at the total read lengths
because many problems can lead to a reduction in the number of high
quality bases while generally the number total bases called by phred
will tend to be less affected.) If your laser is drawing over 9 mA I
would advise extreme vigilance until you can get it replaced.
    Another detail. Last time my laser was cutting out, it seemed to be
happening at random intervals. This time, I'm seeing whole blocks
(window-sized areas of the collected data in the array view pane)
completely blank with the data immediately before and after looking
completely normal. Almost as if some sensor is being checked just before
the pane is refreshed and if the laser is (for example) drawing over a
certain current, it is shut down for exactly one pane to allow it to
cool, and restarted the follwing pane.
    In any case, I'll post more after the service engineer has a chance
to look things over (and probably replace the laser). If anyone has seen
this phenomenon I would appreciate them emailing me or (better) posting
    My impression is that most places with more than 1 or 2 3700s are
likely in the process of upgrading to the 3730, but still there will be
a substantial number of users who are still running 3700s under v.2 Data

Phillip SanMiguel
Purdue Genomics Core Facility


More information about the Autoseq mailing list

Send comments to us at biosci-help [At] net.bio.net