[seqfan] Re: Suggested keyword: veri

franktaw at netscape.net franktaw at netscape.net
Tue Apr 14 21:35:29 CEST 2009


I did quite deliberately say *mostly* a waste of time.  If the 
usefulness of
verifying a sequence can be approximated by the probability that an 
error
will be found, verifying a hand-calculated sequence is 10 to 10000 
times as
useful as verifying a machine-calculated sequence.  Now, if we were 
really
going to verify every sequence, this wouldn't matter; but 
realistically, we
aren't.  So what I want, on those occasions when I do feel like 
verifying,
is some way to find the more useful candidates for verification.

A comment identifying sequences that have been verified quite simply
doesn't do this.  Searching for sequences that don't contain "verified
by" will find most of the database.  I don't know of any effective way
to search for (as opposed to identify by inspection) sequences that
don't have programs associated with them.

On the other hand, only a tiny fraction of hand-calculated sequences 
have
a comment like "should be verified" or "needs verification"; and it is 
not
likely that people will soon start entering such lines in most cases.  
I do
think that people would generally check such a keyword when applicable
if it was available.  This doesn't help with the backlog, but would 
help
going forward.

Franklin T. Adams-Watters


-----Original Message-----
From: Hugo van der Sanden <human at google.com>

2009/4/12 <franktaw at netscape.net>

> Yes, but there are 150,000 sequences there, and only a handful have 
any
> such entry.  The vast majority of them are machine-calculated, and
> checking them is mostly a waste of time. [snip]

I disagree that verifying machine-calculated values is a waste of time:
there may be bugs in the programs we write to calculate; there may be 
bugs
in the maths packages we take advantage of; there may be bugs at the
firmware or hardware level (the well-publicised Intel division bug is 
by no
means the only example).

....

Since we have an existing mechanism to record this that's capable of
capturing all this information, all we need to do is get in the habit of
using it. :)

Hugo




More information about the SeqFan mailing list