Do you have to sequence both strands? (anecdotes)

Fri Apr 15 08:52:29 EST 1994

It can actually get worse than you think.  In the past we have found
sequences in Genbank that had gross errors at either *end* of the
sequence.  Considering where the errors occurred, we suspect that
someone tried to eke out every last bit of sequence that they could out
of the gel ("Gee, those bands are really starting to run together at
the top, but I think it says gaacagata, don't you?").

> Over the past five years my lab has sequenced at least twenty new cDNAs in a 
> competitive area.  We've taken a variety of approaches such as subcloning 
> fragments, nested deletion series (the best in my view), custom oligos; 
> mainly using double-stranded DNA, occasionally using an ABI sequenator.  
> While I wholeheartedly agree in the need to sequence both strands, there are 
> several anecdotal observations I'd like to add.
> One project involved a particularly GC-rich, novel gene (about 3.5 kb).  The 
> student cloned both bovine and human cDNAs and bombarded them with all of the 
> above strategies.  Over twenty specific oligos were used to read across 
> compressions, etc.  Various denaturation techniques were employed.  Both 
> strands in both genes were sequenced yet we still made two separate errors 
> which shifted and then restored the correct ORF over a 40 base pair stretch. 
> Moral: doing everthing "right" doesn't guarentee accuracy.
> Another project involved a gene from slime mould which has an AT-rich genome.  
> We used both radioactive and automated sequencing and the latter was abysmal.  
> ABI used this aprticular gene to specifically improve the 373s performance in 
> dealing with AT repeats, using the radioactively-determined sequence as the 
> benchmark (which resulted, in part, to the introduction of the 5-filter 
> wheel).  Moral: hi-tec isn't always hi-acc.
> This week I downloaded a sequence from NCBI via its accession number.  
> Translation of the cDNA did not yield the same ORF as in the GenBank entry 
> due to a single base deletion.  Although programs like Authorin check for 
> such errors, many submissions are made without these tests. Moral: some 
> mistakes are easily found.
> Finally, in a competitive area, the degree of redundancy of sequencing has to 
> be kept to a minimum for reasons of both efficiency and effectiveness (i.e. 
> each strand should be sequenced only once).  This is perhaps becoming more of 
> a problem.  Short cuts are bound to be taken.  How many people sequence both 
> strands (or even entirely one strand) of a site-directed mutant?  There are 
> controls (protein expression, etc) but I've certainly chanced upon unintended 
> errors which would have been undetectable other than by sequencing.  Moral: 
> Cell/Nature/Science don't wait for ever.
> Enough moralising, got to get this paper off.......
> Jim

James F. George, Ph.D.              "Back off man, I'm a scientist"
Department of Surgery                --Bill Murray
University of Alabama at Birmingham
205-934-4261 voice
txpljfg at

More information about the Methods mailing list