anyone looked at long-range patterns in DNA sequence?

suresh krishnajois sureshk at
Tue Mar 23 19:48:13 EST 1993

In article <1oh01iINNh0f at>,
xia at (Xuhua Xia) writes:
|> In article <1993Mar17.153421.21757 at>
mlevin at (Michael Levin) writes:

...[deleted stuff]....
|> >
|> > What tools (mathematical) can be used to look for patterns
|> >in a string of symbols (when the pattern isn't known in advance)?
|> >Please send ideas to mlevin at husc8.harvard.eddu.
|> >
|> >Mike Levin
|> >
|> You can use time series with different time (base) lags. What pattern
|> are you suspecting? It sounds pretty wild.
|> Xuhua Xia

Try using neural networks with unsupervised learning. Other math tools include
cluster analysis. 

I am not a practicing bio-phycisist, but I believe there are three MAJOR 
problems with long-range pattern analysis:

1. parametrization of the pattern space (i.e., to look for patterns of what ?)
2. scale of pattern recognition (i.e., once the pattern space is chosen,
at what 
   physical / biological / mathematically plausible scales do we choose to look
   for patterns ?
3. is a static structural analysis good enough ? Is pattern analysis of DNA
   sequences any more (or less) useful than trying to understand a software
   product just by reading its source code ?

In an earlier post on this thread, I had mentioned that me and a bio-phycisist 
colleague had tried to apply dynamical system theory to DNA sequeces.
The results
were intriguing, but inconclusive. Also, I felt the approach was conceptually
inadequate due to the three reasons I have mentioned above. I am trying to
work this out both bottom up (low level details) and top down (the above three
conceptual problems). May be other netters in this and related news groups
could contribute their thoughts to this fundamental issue.

- Suresh

Reply To:    Suresh Krishnajois
email   :    sureshk at
Phone   :    415-390-5454 (work)
Methinks:    I think, therefore my head aches  };-(

More information about the Comp-bio mailing list