LOD score of a single genotype

Question

[rbagnall]

Hi,

I have a family pedigree that shows autosomal dominant inheritance of a trait. I have sequenced a know causal gene for the phenotype and find a novel missense variant. Genotyping the missense variant in the family shows that it is present in all 6 affected family members and absent in 4 brothers/sisters and offsping of the affected members. It is also absent in 4 parents who married into the family. Using Superlink software, the LOD score based on genotyping the single variant is 2.98.

1. Is this LOD score equivalent to doing a genome wide linkage analysis in the family with a marker segregating with the disease at this locus?
2. Are the 4 people who married into the family as informative to the LOD calculation as the 4 unaffected brothers/sisters and offspring. Should I even be including the 4 people who married into the family in the LOD score?

Or is it not correct to calculate a LOD score for a single genotyped variant?

Thanks.

Answer 1

[Lemire]

It's been over 10 years I dealt with linkage, but let me have a try...

To illustrate, I downloaded the test examples (datafile.dat and pedfile.dat) at http://bioinfo.cs.technion.ac.il/superlink/

I modified the last two line of datafile.dat:

0.0000 << RECOMBINATION VALUES
1 0.0500 0.50000 << REC VARIED, INCREMENT, FINISHING VALUE

Superlink will do two-point linkage: it will compute the likelihood of the data for each of the recombination values (thetas) specified in the two lines above, between your marker and the trait locus (i.e. recombination values ranging from 0 to 0.50 in 0.05 increments, as specified).

In the output you'll get a series of tables (one per theta) such as

Loci order: 1 2
THETAS  0.100 
--------------------------------------------------
PEDIGREE |  LN(LIKE)  |  LOG 10(LIKE) | LOD_SCORE
--------------------------------------------------
   1      -173.338440   -75.279928      -0.402989
--------------------------------------------------
TOTAL     -173.338440   -75.279928      -0.402989    

-2 LN(LIKE) = 346.676880

Now this LOD score is for the test H0:theta=0.10 vs H1:theta=0.50. The rest is just maximum likelihood, you go fetch the theta estimate with maximum likelihood, and use/report the corresponding LOD. The final table somewhat summarizes it all, except that it reports cM instead of thetas,

--------------------------------------------------------------
  Marker | Trait Position (cM) |   LN(LIKE)  | LOD_SCORE
   Name  |  (from marker #1)   |             |
--------------------------------------------------------------
                -1000.000000     -172.410523    -0.000000
                -115.129255      -172.420514    -0.004339
                -80.471896       -172.450027    -0.017157
                -60.198640       -172.505459    -0.041230
                -45.814537       -172.592764    -0.079146
                -34.657359       -172.717451    -0.133297
                -25.541281       -172.884010    -0.205633
                -17.833747       -173.093725    -0.296711
                -11.157178       -173.338440    -0.402989
                -5.268026        -173.586474    -0.510709
  loc1
                0.000000         -173.765325    -0.588383
--------------------------------------------------------------
  MAX           UNLINKED         -172.410523    0.000000
--------------------------------------------------------------

Here the LOD derived from maximum likelihood is 0 corresponding to an unlinked marker.

So, to answer your questions:

1. it's ok to use just one marker, there might be enough information to estimate the recombination fractions. Following what I described, the LOD you report would be one obtained from maximum likelihood, which is not the same as the genome-wide setting that tests theta=0 (i.e. the marker is the trait locus, if that is what the genome-wide setting is testing, I am unsure about that).

2. it's ok to leave more individuals in the pedigree, if these are truly uninformative they won't have an effect on the derivation of the likelihood.

[rbagnall]

Thanks Lemire for taking the time to answer my question, much appreciated.

Ultimately, I think I need to sit with someone who can walk me through my data and explain things step by step.

[GabrielMontenegro]

This is not completely related to the question, but in the case were you have different families, why is it possible to just add up the LOD-scores over different pedigrees? Thanks!