Typically a P-value of p>0.05 should be applied for HWE, although in certain instances is might be useful to apply p>0.005. This may help generate values approaching 1 instead of value = 1 which would skew the results.

It is important to note that the "scale" of HWE lies between Zero and One and thus any SNP approaching 1 will impact on the result. Furthermore, it is a measure of the genetic/allelic continuity of a specific population and, as such, is useful in determinig how specific mutation come about and their source.

Your results suggest that you've identified these SNPs in a single population, and that they are not consistent with what's published in databases such as HapMap. This means that although slightly "biased", the HWE value you generate is a reflection of specific SNPs in an isolated population. Moreover, it would then be safe to assume that since it is an isolated ppopulation, the "system" is in HWE, taking into account the age and sex of your sample population.

De novo mutations, selection, genetic drift, gene flow, meiotic drive and mate choice can all theoretically bias the allele and genotype frequencies and thus the HWE. A difference in allele frequency in a population, comparing to the reference can be shown for each individual allele with the help of simple statistical tests. For example, I believe that a two-tailed Fishers exact test (or Chi-square for large samples) could be used in an attempt to refute the null hypothesis that an allele has the same frequency in the studied population and reference, assuming observations are independant. You have n out of N cases of variant A in the studied population and m out of M cases of variant A in the reference population. This forms a 2x2 table. If the p-value is less than threshold (

So, if the reference group is AA and I found for AA 98.2% and for AG I found 1.8% (n = 1) it would be alright to compare these two? Moreover, around 1/6 of all single-nucleotide  variant is a mutation consequently shown such an frequency. In this case, would you advise to leave these out of the equation,  keeping only those with a frequency of at least > 0.01? 

Hi Rene, I am a little unsure where 1 in 6 is coming from?  The likelihood of a mutation at one particular base in any given individual is remote.  The mutation rate in humans is estimated at 1 change per 100 million to 1 billion bases/generation.  The likelihood of any SNP you observe in a human genome being a de novo mutation is quite small.  The only way to be sure is to sequence the parents for the same region.  A Chi Square test can be used to answer your question, but you will use the actual numbers, not the percentages, to do the calculation.  Also, in order to compare populations, you have to be able to define them.  A population of 100 volunteers is not sufficient, so be sure to consider carefully if you have tested a population in a biologically and statistically defined and meaningful manner.  Hope that helps!

Hi Rene, if you feel that prevalence of snps are low, you could set a maf cutoff like 0.05, and exclude snps with maf

Dear Linda,

Thanks for the link. Unfortunately, I dont really understand how to use the program.

Do you have any experience?

Kind regards,

René

Dear Linda,

I did try (see link below), but not sure how to interpret the results:

http://ihg.gsf.de/cgi-bin/hw/hwa2.pl

Kind regards,

René

Click on the blue 'SNP' hyperlinks for each snp to open a new page showing HW hyperbola. Significant p-values and HW deviations are shown in red.

For the above mentioned SNP I only got a blue and a black dotted line depicted in the Finetti diagram. So, this means no departure of HWE right?

I'm not really sure I understand your question but it seems like you have a very small sample and you want to know if it is in HWE. First off, you should check to see how your frequency compares in reference databases in a similar population, eg on dbSNP. Second, if the frequencies are different, I would check that you haven't made some mistakes -  e.g. genotyping errors, or data transcription errors. Genotyping errors are the most likely explanation for deviation from HWE. I am guessing you have AA = 54, AG = 1, and GG=0. You could do a Fisher's exact test to see if your observed distribution is significantly different from your expected distribution. 

Dear Philip,

Are you available for some assistance today?

Kind regards, 

René

Another question, HAPMAP uses chrom. sample count whereas I use the number of patients. Should I multiply the number of patients by 2 to get a more realistic view of the true number of genotypes?

Hi Rene, The hyperbola is graphed based on the frequency of alleles.

Legend:

- Length of vertical line: frequency of genotype 12.

- Length of left perpendicular line: frequency of genotype 11.

- Length of right perpendicular line: frequency of genotype 22.

- X-axis: frequency of allele 1.

- Intersection of Hardy-Weinberg

parabola and vertical line: frequency of genotype 12 in case of Hardy-Weinberg equilibrium.

The dotted line of the hyperbola signifies the points at which the HWE holds, if either of the three possible genotyoes of a bi-allelic marker strays too far out of the hyperbola, then it suggests deviation from HWE. The blue dotted line represents cases and black represents controls. Just try varying the genotype numbers and watch how the HWE change.

Calculate your minor allele frequency  AG+GG*2/ (AA+AG+GG)= 1/110 = 0.009. What's the minor allele frequency in dbSNP? How did you do the genotyping? 

Dear Philip,

The MAF varies between your number and 0.25. I am not sure what you mean by how I did do genotyping. I sequenced the whole gene and compared the result with a reference sequence. Nex I compared these results to the NCBI database.

Dear Philip,

Is there a minimal MAF to HWE calculation? 

genotyping method can affect interpretation of results. Some methods are much more sensitive to genotyping error and are unrealiable for rare SNPs. What kind of sequencing? was it a capillary sequencer? 

What population is your sample from? What is the MAF in dbSNP for the population that matches your population? There is no minimal MAF for HWE calculation. For HWE, you're just comparing observed to expected number of genotypes. 

I used capillary sequencing in both forward and complement.

The population consists of European female volunteers that were previously diagnosed with a multicausal bleeding disorder. 

As for the comparison to controls, it still feels strange to compare casus and controls that just have 1 observation eg MAF 0.02. 

Furthermore,  If the HAPMAP SNP (MAF > 0.01) shows deparute from HWE, can I still use this as control?

Dear Rene

sorry If I answer now.

I have experience with 54K chip for bovine. When my statistical colleague have analized the results of genotyping using this chip on bovine DNA (more 3000 samples) they have excluded SNP with a MAF < 0,01.

Dear Cinzia, 

Do you have experimence with performing and interpreting HWE data?

Kind regards,

René

I am sure that until now you have done your analysis. This is a useful aricle for new readers.

Article New alterations at potentially regulated regions of the Glia...

If working on a single SNP having HWE > 0.05, how to explain it, as it is the only targeted SNP.