Ldheatmap Does Not Include All Snps In The Analysis (Tri-Allelic Snp Excluded)

Question

[Agatha]

I am trying to generate an LD plot using LDheatmap in R.

I have 4 SNPs geotyped for around 80 000 individuals :

head(m.genos)

   p1205 p0864 p3077 p1244
1     G/G     C/C     T/T     C/C
2     A/A     C/C     T/T     C/C
3     G/G     T/T     T/T     T/T
4     A/A     C/C     T/T     C/C
5     G/G     C/C     T/T     C/C
6     G/A     C/C     T/T     C/C

MyHeatmap <- LDheatmap(m.genos, genetic.distances = as.numeric(m.dist),distances="physical",LDmeasure="r",title="LD analysis",    
                       add.map=TRUE,add.key=TRUE,flip=TRUE,SNP.name=names[1:4])

However, the LDheatmap function computes the ld matrix using ONLY 3 of them.

> MyHeatmap$LDmatrix
           p3077    p1205     p0864
p3077      NA 0.02379055 0.02184959
p1205      NA         NA 0.21808013
p0864      NA         NA         NA

I have managed to succesfully generate the LD matrix for other SNPs, so I was wondering what could me the problem in this case?

The excluded SNP is tri allelic : => possibly causing the LDheatmap function to fail in generating the complete LD matrix

  Number of samples typed: 87058 (99.8%)

  Allele Frequency: (3 alleles)
  Count Proportion
 C  110853       0.64
 T   54539       0.31
 A    8724       0.05
 NA    272         NA


 Genotype Frequency:
    Count Proportion
   C/C 35354       0.41
   C/T 34647       0.40
   C/A  5498       0.06
   T/T  8552       0.10
   T/A  2788       0.03
   A/A   219       0.00
    NA    136         NA

  Heterozygosity (Hu)  = 0.4940388
  Poly. Inf. Content   = 0.4119684

Answer 1

[Fabio Marroni]

I think that several tools for calculating LD for SNPs will exclude triallelic SNPs (because I think they make computation somewhat more difficult). If you use the r package genetics, for example (which is the one used by LDheatmap for the calculation, if you let it do the calc) it will remove triallelic SNPs. An option might be to use tools that calculate LD for multiallelic markers. If you only have 4 SNPs it should not be difficult. For example, Arlequin can do that. I guess many other tools can do that.

[Agatha]

that's probably it...thought so as well. thanks for the alternative suggestions.