multipoint runs

Mon Jun 26 02:07:37 EST 1995

I was interested in the following question posted on 6/13/95:

> We are mapping a disease gene.  We have four markers (A, B, C, and D)
> that are linked to the gene.  We ran two mutipoints: first, phenotype and
> A, B, and C; second, phenotype and B, C, and D.  Now here is my question:
> Does anyone knows the best and the right way to put these two runs together.
> Alex Nechiporuk,  Cedars Sinai Medical Center, Los Angeles.

He subsequently wrote:

> I want to thank everybody who posted answers on my yesterday question
> about multipoint runs.

I saw no posted answers except the two from Dave Curtis.  Where other
suggestions/answers posted?  If they were sent directly to the 
original poster I would appreciate if they could be posted to the group or
forwarded to me.

In the example previously presented for this situation, 
the location scores are not very different in the two runs for the 
overlapping interval (i.e. in the interval between markers 12 and 10). 
If you look at a ratio of the location scores in this interval the ratio
ranges from .95 to 1.02.  Is it valid to compute an odds for order, 
for a particular location of the disease gene in the second run, say
12--------10=====1======11   loc. score 
   .068     .013   .053         49.67

using the max location score in the first run, which was the overall
max in the two runs?  I realize that Dave Curtis said 

> It can't be right to use either the highest or the lowest as estimates 
> of the "true" multipoint lod score.

And I agree.  I wondered though if it would still be misleading or incorrect 
to plot the results in a single line in this case, when the scores in the
overlapping region were so similar across the two runs.

If the location scores were less similar between the two runs, one could 
scale the location scores from run one by:
1) obtaining the ratio of the loc scores in run one to run two in the 
overlapping interval 
2) multiplying the loc scores of run two by the ratio (or even an average 
  ratio as the ratio may differ within the overlapping interval) 

One could then compute odds for order for all points in the new
combined run.

Note I said one *could* do this, I am not suggesting it is necessarily 
the right thing to do.  I would appreciate any comments or feedback 
about the scaling process described.

Anita DeStefano, Ph.D.
Boston University Medical Center           
Dept. of Neurology L320            TEL: 617 638-4057
80 E Concord St.                   FAX: 617 638-4275
Boston, MA 02218-2394              email: destefano at

More information about the Gen-link mailing list