LOD score too high - answers

Daniel E. Weeks dweeks at WATSON.HGEN.PITT.EDU
Thu Feb 9 14:59:25 EST 1995

On Thu, 9 Feb 1995, David Curtis wrote:

> To test the hypothesis of all linked against none 
> linked you'd have 1 - 0.0 - 2 - 0.0 - 3 against 1 - 0.5 - 2 - 0.5 - 3, which 
> would mean changing two intervals simultaneously. 

   Unfortunately, this is exactly what LINKMAP does when it computes 
Ott's generalized lod score (which is described in his book).  I think 
this may be identified only as a lod score in the LRP output, but in the 
final.out text output, it is identified as Ott's generalized lod score 
(without further explanation).  LINKMAP should compute multipoint lod 
scores, but it cannot since it only does one order at a time.  

For example, if 1 is the trait locus, and our given marker map is 
2---0.05---3, then the multipoint lod score for the order 
2---0.0---1---0.05---3 requires computation of two likelihoods:

A) The likelihood under order 2---0.0---1---0.05---3
B) The likelihood under order 1---0.5---2---0.05---3 (i.e., with the 
trait off the map).

Thus, post-processing of LINKMAP output is required to compute valid 
multipoint lod scores.  I believe Jurg Ott distributes a utility 
program to do this.  I have one as well, which is now available in the 
file 'map.tar.Z' by anonymous ftp from watson.hgen.pitt.edu.  Please note 
that my conversion program requires that you start each group of LINKMAP 
runs with the trait locus off the marker map on the left.

I hope this helps clarify matters.

-- Dan Weeks --
 Daniel E. Weeks, Ph.D.        
 The Wellcome Trust Centre     Department of Human Genetics   
 for Human Genetics            University of Pittsburgh        
 University of Oxford          Crabtree Hall, Room A310     
 Windmill Road                 130 DeSoto Street             
 Oxford OX3 7BN                Pittsburgh, PA 15261          
 (+44) 865 742 441             (1-412) 624-3066               
 FAX: (+44) 865 742 196        FAX: (1-412) 624-3020          
 daniel.weeks at well.ox.ac.uk    dweeks at watson.hgen.pitt.edu  

More information about the Gen-link mailing list