A SMALL COMPARISMENT OF LAMP, ADMIXTURE AND AIMs

INTRODUCTION

The LAMP 2.5 genetic analysis software picks AIM's in its analysis based on the allele frequences given to it. On this basis I would like to check how LAMP perform by comparing to the widely used ADMIXTURE software. Of special interest are if ADMIXTURE clusters individuals similar to LAMP.

METHOD

I first estimatet the ancestral frequency data for Norwegians, Swedes, Saami and Finns from a subset of individuals clustering to these based on their MDS plots. I then run LAMP on recommended settings except this time I run it assuming the populations had been mixing for 10 generations the same presumption used in ADMIXTURE. I then picked the AIM's chosen for this analysis by LAMP and used them in a run for the whole dataset of 36 individuals.

OBSERVATIONS:

* All the Finns who clustered close to or 100% in LAMP also did the same in ADMIXTURE.

* In LAMP 9 Norwegians appeared 100% or close to 100% Norwegian. In ADMIXTURE 2 of these 9 appeared strongly mixed with Saami and Finns in ADMIXTURE.

* Saami all appear in LAMP close to or at 100% Saami. In ADMIXTURE only 1 appears as heavily mixed.

* All the Swedes who appeared as 100% Swedish in LAMP was also classified as 100% Swedish in ADMIXTURE.

* The Norwegian mixture in Swedes and Finns appears stronger in LAMP than in ADMIXTURE for those individuals who have it.

* Saami amix proportions seen in considerabe proportions in ADMIXTURE are often very little or gone in LAMP.

COMMENT:

It seems LAMP and ADMIXTURE is largely consistent about the "pure breed" and supports that the ancestral allele frequency data used in LAMP are correct. What LAMP and ADMIXTURE do disagree some about are the admixtures. 3 individuals who in ADMIXTURE was classified as mixed was not or very little admixted in LAMP. The reason for this are that LAMP and ADMIXTURE uses different methods to estimate admixture.

ADMIXTURE basically estimate allele freqency differences between SNP that is assumed to be independent and than cluster it. It does not care about physical positions of SNP, cM (centimorgans) between SNP and recombination events as far as I can tell.

LAMP however first picks AIMs (ancestry informmative markers) and goes trough the whole chromosome piece by piece in a moving window that changes in size depending on the physical position and the recombination rate of the window where it constantly compares to the given ancestral alleles for different populations. In this it differ from ADMIXTURE by looking at a group of SNP instead of a single SNP as ADMIXTURE. It comes without saying that comparing a group SNP gives a higher likelyhood of differentiating the actual origin of a segment than a single SNP.

This likely explain why some individuals that showed considerable admixture in ADMIXTURE had little or no such admixture in LAMP and why some proportions of other admixtures seen in ADMIXTURE was gone in LAMP. If looking only at single SNP's and averaging them some people may have f.ex "Saami" looking allele frequencies spread over the chromosome but when checked vs a segment of SNP's in the same segment they where actually more likely not of Saami origin.

In ADMIXTURE such problems as mention above go unnoticed unless the admixture is real and you can actually observe by comparing ADMIXTURE vs LAMP results that where the results appear consistent in both analysis its probably real or at least consistent given the parameters and data. This is because if the admixture is real a considerable of number of single SNP's with frequency close to a source populations will be there and it will of course by detected by ADMIXTURE.

Anders