home   genetic news   bioinformatics   biotechnology   literature   journals   ethics   positions   events   sitemap
 
  HUM-MOLGEN -> mail archive   |   Search register for news alert (free)  
  Jurg Ott: LITE  
   

archive of HUM-MOLGEN mails

 
 

[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Topic Index]

To: Multiple recipients of list HUM-MOLGEN <HUM-MOLGEN@NIC.SURFNET.NL>
Subject: LITE
From: Jurg Ott <ott@linkage.cpmc.columbia.edu>
Date: Thu, 12 Dec 1996 15:28:15 -0500

                            Linkage Newsletter
                     Vol. 10   No. 3   December 1996

Published by Jurg Ott                     Postal address:
  Email: ott@rockvax.rockefeller.edu      Rockefeller University, Box 192
  WWW:   http://linkage.rockefeller.edu   1230 York Avenue
                                          New York, NY 10021-6399

Editorial Assistant:  Katherine Montague  Fax: 212-327-7996
  Email: montagk@rockvax.rockefeller.edu  Tel: 212-327-7979

(this and all previous newsletters are available on our ftp site/Web page)


EDITORIAL

   My collaborators and I have now completed our move to Rockefeller 
University (see new telephone and fax numbers above), where we form the new 
Laboratory of Statistical Genetics. Our web page 
(http://linkage.rockefeller.edu) and ftp site (linkage.rockefeller.edu) are
up and running. We welcome visitors to spend time with us - such interactions 
tend to be very fruitful.


LINKAGE COURSES

   The next linkage courses will be held as indicated below. Interested 
researchers are urged to apply early as our courses tend to fill up quickly. 
Dates for advanced courses (fall of 1997) are not yet known (advanced courses 
are reserved for individuals with a very good background in linkage analysis).

   March 10-14, 1997, at University of Zurich, Switzerland (basic course, 
maximum of 18 participants).  Deadline for applications is January 15, 1997, 
which is when applicants will be assigned space in the course; applications 
submitted after the deadline will still be considered but have a much smaller 
chance of success.

   June 23-27, 1997, at Rockefeller University (basic course, maximum of 30 
participants).  Deadline for applications is May 3, 1997.

   To obtain information on these courses, please write to Katherine 
Montague, course coordinator, by email (preferred) or fax.

   We will use our book (Terwilliger and Ott, Handbook of Human Genetic 
Linkage, Johns Hopkins University Press, 1994), with supplemental handouts for
advanced courses.  Participants are expected to buy the book and bring it to 
the course;  in case of problems please contact Katherine Montague in advance 
of the course.  A list of corrections for the book may be downloaded from our 
anonymous ftp site, linkage.rockefeller.edu (file corr_ter.txt in directory 
book).


SOFTWARE

>> ftp sites <<

Please note, as mentioned above, that our current ftp site is

   linkage.rockefeller.edu

Our previous site, linkage.cpmc.columbia.edu, is now being used by Dr. Joseph 
Terwilliger at Columbia University, where he is making his programs available 
for distribution.


>> Bug Report <<
(Dr. Daniel Weeks, University of Pittsburgh and Wellcome Genome Center, 
 Oxford)
                                                           27 Nov 1996
                FASTSLINK and SimIBD bug report

Thanks to the help of Hakan Sakul at Sequana, we have tracked down a
subtle bug in FASTSLINK (which was due to a one letter typo).  The source
code on the ftp server watson.hgen.pitt.edu has been corrected, so you can
just re-ftp the software from there (or from the UK mirror site 
ftp://ftp.ebi.ac.uk/pub/software/linkage_and_mapping/hgen_pitt/).

Since SimIBD uses FASTSLINK to do the simulations, SimIBD is influenced by
this bug too.  Please re-ftp the SimIBD package also if you use it. 

NOTE: We have developed a diagnostic version of FASTSLINK which allows you
to easily check if your pedigrees were such that your results would have
been influenced by this bug. 

MANUAL CORRECTION OF FASTSLINK CODE 

To correct the FASTSLINK code itself, instead of ftp'ing the whole
package, you may just correct the one line that is in error as follows:

The bug is a typo at line 741 of slautomodified.c.

The current incorrect code reads:
parunk = (((*LINK->p)->geneloc == 0) || ((*LINK->p)->geneloc == 0) || !noloop);

It should read:
parunk = (((*LINK->p)->geneloc == 0) || ((*LINK->q)->geneloc == 0) || !noloop);

EFFECT OF THE BUG

The error causes the segdown routine to do the wrong thing under some
circumstances. At first glance this looks like a bug that should have
shown up a lot sooner. However, it won't show up in FASTSLINK if: 

1. Spouses always get consecutive numbers.
   or
2. The pedigree has a loop
   or
3. The pedigree is simple and numbered top to bottom in increasing
   order (i.e., children always have ids larger than their parents). 

The problematic line of code is trying to determine if both parents have
already had their genotype selected. In segdown, one updates probabilities
of a child based on parents and siblings. If both parents have their
genotype known, then the siblings can be ignored. This is advantageous
because the code for the one child case (whether in the original or fast
versions) is faster than for the many child case. 

In segdown p is always the father and q is always the mother.

The problem arose only in the loopless case.  The effect of the typo was
to go into the one child case if the father already had a genotype
assigned. This would be erroneous if the mother did not have a genotype
yet. 

The effect of the error depends on the random seed, which is why Hakan
Sakul's pedigree showed different behavior with different seeds. 

In some cases, ignoring the effect of the siblings causes an incompatible
genotype to be assigned, which would eventually lead to a crash. 

In other cases, a compatible-with-siblings genotype would be assigned by
chance and the pedigree replicates would come out compatible, but sampled
from a wrong distribution, which is extremely unfortunate. 

Because FASTSLINK fills in genotypes in increasing order of ids, the only
way the problematic code could have an effect is if there is a nuclear
family where the father has a lower id than the mother.  BUT, the pedigree
traversal algorithm ensures that the call to fill in the proband is a
segup call, not segdown. Therefore, if mates always have consecutive
numbers each traversal has at most one nuclear family with father filled
in and mother not AND the mother is the proband for that call so she gets
filled in in segup and not segdown. 

If the pedigree is simple and numbered from top to bottom, segdown is
never used. 

For SimIBD, some people are skipped when simulating, changing the normal
order in which slink does the simulations.  Therefore, even in a nuclear
family, if p is the mother and if the children are affected but the mother
is unaffected (and skipped), then the bug is hit.  However, if the father
is unaffected, then the bug is not hit.  Therefore, because SimIBD forces
FASTSLINK to do the calculations in a different order from the original
FASTSLINK, SimIBD results might be affected in other situations as well.

A DIAGNOSTIC VERSION

We have constructed a diagnostic version of FASTSLINK, which you can make
by typing

make slinkdiag

Then you can run the diagnostic version on each pedigree you used to see
if any would hit the bug. To do this, set up as you would for a normal
SLINK run, but type 'slinkdiag' instead of 'slink'.  Note that it is only
necessary to generate 1 replicate per pedigree to test for the effect of
the bug.

If the bug is hit, then you will see a message appear on the screen like:

BUG WAS HIT: Pedigree 3  Person 255


>> ANSI.SYS driver in Windows 95 and NT <<

   We keep receiving requests for support because "the LCP program produces 
only gibberish on the screen". The reason is a missing ansi.sys driver and 
not, as some users concluded, that the programs don’t run under Windows 95 or 
NT. Currently, to run the LINKAGE programs under Windows 95 or NT, a DOS 
window must be opened and the programs run there. We are in the process of 
adapting them to Windows NT using the corresponding new NDP Pascal compiler. 
Readers of this newsletter will be notified when work is completed.

   Windows 95: As is mentioned in the msdosdrv.txt file in the windows 
directory, the ansi.sys driver must be installed by inserting a line in the 
config.sys file as follows:
   device=c:\windows\command\ansi.sys
The config.sys file is in the root (top) directory and may initially be empty.

   Windows NT: In version 3.51 of NT, you open the Control Panel, double 
click on System, and click on Environment. Then, you define a new variable, 
device, and give it the value c:\winnt351\system32\ansi.sys (this is how I 
remember it). In version 4.0, according to the documentation, the config.nt 
file must be changed in a similar manner as for Windows 95.


>> Computer program for identification of relationship errors
   in sib pair studies <<
   (Harald H. Goring)

In affected sib pair studies, the power to detect linkage will generally be 
decreased if non-sibs, which were erroneously believed to be sibs, are
present in the collected families. While such relationship errors will often
be detected through mendelian inconsistencies, this is typically not the case
in studies of late onset diseases when the parents of affected individuals
are already deceased. To detect non-sibs also in this situation, I wrote a 
computer program called "relative". A paper describing the method has been 
submitted for publication and an abstract has been published: Göring HHH &
Ott J (1995) Verification of sib relationship without knowledge of parental 
genotypes. Am J Hum Genet 57, A192.
 
The "relative" program computes probabilities that two stated sibs are in 
reality unrelated, half-sibs or sibs. This is accomplished by computing the 
likelihoods for the three relationships from the observed genotypes of the
two individuals. For this estimation, the same genotypes may be used that
were obtained by marker typing for linkage analysis. The more markers are
available, the better the estimation procedure performs. Linked markers are
handled correctly by the program, and a single parent may be typed alone. In
addition to standard LINKAGE-format input files, the user must supply prior
probabilities for relationship errors. The program then identifies in the
pedigree file all typed sib pairs that do not have both parents typed and
carries out computations for each sib pair in turn. After non-sibs are
identified by the program, the investigator may remove them from the sample
of collected families. This tends to increase power for linkage detection.
Note that the program can also compute likelihoods for some other common
relationships.

The "relative" program is distributed on our ftp site
(linkage.rockefeller.edu) in directory /software/relative.


>> Request for Information <<
   (Dr. Wentian Li)

Under the web page of our laboratory at Rockefeller University, there is an 
"alphabetic list of genetic analysis software" which lists more than 110 
linkage analysis, genetic map construction, and pedigree drawing programs.
The URL (uniform resource locator) is:
 
   http://linkage.rockefeller.edu/soft/list.html
 
For each program, I try to include information on author, ftp site, web home 
page, source code language, references, etc. For many programs I cannot find 
all of this information. To keep the list up-to-date, it would be very
helpful if the authors of these programs could check the list and send us
any missing information, missing links, and new release information (please
send email to wli@linkage.rockefeller.edu).
 
The current format is the following:

1.  Full program name: e.g. if you have a program called APM, identify it as 
    "Affected Pedigree Member (method)".
2.  Version: e.g. v3.8 (Dec 1996)
3.  Description: any description of what the program does.
4.  Authors: ...
5.  Web: home page if any.
6.  FTP: ftp site if any.
7.  Source code language: e.g. Fortran, C, C++...
8.  Operating systems: e.g. DOS, Windows 95, Windows NT, UNIX (SunOS, OSF1). 
    This list need not be complete, but at least list the operating systems
    on which the program is known to be working.
9.  Executables: if the executable name is different from the program name
    it is advantageous to have these names.
10.  References: major publications relevant to the program such as the
     program announcement or the theoretical paper on which the program is
     based.
11.  Price: if it is a commercial package.
 
A more up-to-date list will benefit everybody, and your help will surely be 
appreciated.


>> Genehunter update - version 1.1 <<
 
Several changes have been made to the Genehunter software and are included in 
version 1.1 - currently available via anonymous ftp at genome.wi.mit.edu in 
distribution/software/genehunter.
 
* We implemented an algorithmic improvement which should increase analysis 
speed by a factor of 5-10 in many cases.  The speed-up was motivated by the 
work of Ramana Idury, Genetic Epidemiology 13:A311, 1996.
 
* We implemented a fix to allow compilation under AIX.

In addition, users should note that the use of the KOSAMBI map function can 
produce slight numerical errors in LOD score values in intervals between 
markers. We recommend that the HALDANE map function, which is multilocus 
feasible, be used for all analyses and that distances between markers be
input as recombination fractions or in Haldane centiMorgans.  Thanks to Dan
Weeks for bringing this to our attention in an example where a pedigree with
all members designated "phenotype unknown" produced a LOD score of 0.03
(instead of 0) between two markers.

The Haldane map function is the default in the new version of Genehunter. 
Users of the old version, in which Kosambi is the default, should download
the new version as soon as possible.  In the meantime, they are advised to
use the Haldane map function in all analyses by issuing the 'map function
haldane' command when the program is started up.

Information regarding Genehunter can be obtained at the above ftp site or by 
reading our article in the American Journal of Human Genetics (58:1347-1363 
June 1996).  If you have problems or questions regarding the software contact:
 
Leonid Kruglyak  (leonid@genome.wi.mit.edu)
Mark Daly        (mjdaly@genome.wi.mit.edu)

----------------------------------------------------------------------------
Support through grant HG00008 from the National Center for Human Genome 
Research is gratefully acknowledged.


   
 
home   genetic news   bioinformatics   biotechnology   literature   journals   ethics   positions   events   sitemap
 
 
 

Mail converted by MHonArc 2.4.4
WWW: Kai Garlipp, Frank S. Zollmann.
7.0 © 1995-2001 HUM-MOLGEN. All rights reserved. Liability and Copyright.