Hi all: Thanks to all who give me theirs sonsideration about the HWE problem. A lot of people ask me about the answers, here I send some of them... Regards Rodrigo If you have microsat loci not showing H-W proportions you should be aware whether you have a relatively high amount of not-detected alleles ("null-alleles)giving you a surplus of homozygotes. Bo yes, this can be a problem. It depends on why they are out of HWE, how much, and in what direction. Null alleles, Wahlund effect, (selection - unlikely on more than a very few loci)? There is quite a lot of literature on this. You might have seen the recent Evoldir posting about null alleles (question raised by Mike Ritchie) that covered a lot of it, including Selkoe & Toonen (2006). Paul If you use software such as Cervus or Famoz for parentage analysis you can take into account the H-W disequilibrium. These softwares are able work with a level of H-W disequilibrium you can choose. I enjoy to see all the answers you will receive. Pascal I think it will depend on which kind of parentage analysis are you doing. If you are using full exclusion I do not see why it should affect. If you are using a likelihood approach it will propably affect but it might not be to much trouble if the departure from HW is not big. Miguel I think it depend on why they aren't in HWE. One reason might be null alleles. If this is the case, doing paternity/maternity analysis might be quite challenging. On the other hand, if some loci are out of HWE due to more biological causes (migration etc., ) then there shouldn't be much of a problem. Can you tell if you have more or fewer than the expected number of heterozygotes? Too few can be a sign of null alleles. For other questions like estimating gene flow, many people include a locus or two that are out of HWE. I'm not sure this is a good practice, but many of us have chosen this path, rather eliminating the locus from the data set. Jeff Absolutely! H-W equilibrium is certainly an assumption of most parentage determination methods. While I'm not an expert on all the details of each parentage estimation method, I know for a fact that H-W genotype frequencies are assumed for exclusion analyses. Violation of H-W means that the frequency of genotypes is not just the product of the allele frequencies. Usually homozygotes are more frequent and heterozygotes less frequenct than expected under H-W. I work with a tree species where individuals self about 10% of the time. Therefore mating is not random and H-W should not be met exactly. Null alleles at microsatellite loci can also lead to violation of H-W. In principle, it is possible to adjust expected genotype frequencies for increased autozygosity but I am unaware of any program that does this automatically. You should also think about the consequences of violation of the H-W assumption when interpreting your data even if you cannot adjust for it exactly. Good luck, Matt You should probably start by reading: Hoffman JI, and Amos W (2005) Microsatellite genotyping errors: detection approaches, common sources and consequences for paternal exclusion. Molecular Ecology 14, 599-612. A,thoug it doesn't really touch on H-W equilibrium problems it is possible that there might be difficulties. Nobody has really assessed this problem in full. If your interest is paternity or genealogy construction you could look up Fernandez and Toro 2006 manuscript on a new software for genealogy construction that doesn't actually make assumptions of HW. Daniel En el caso de que la información molecular sea "abundante" y muy informativa (valga la redundancia) no habría problema porque a lo que se reduce el análisis de paternidad es a buscar las incompatibilidades e ir excluyendo todos las parejas de padre y madre que no pueden ser y quedarnos con una sola pareja compatible. Trabajando con unos datos de dorada que se genotipaban para 11 micros encontramos casos que no tenían padres compatibles (nos confesó luego la empresa que en algún momento habían tenido unos reproductores que vendieron antes de que se pudieran tomar muestras), pero no encontramos individuos que pudieran asignarse a dos parejas diferentes. Si te salen varios padres compatibles tampoco hay problema porque a lo que se reduce es a calcular la probabilidad de que unos genotipos OBSERVADOS de los padres den unos genotipos OBSERVADOS de los hijos y así puedes ordenar las probabilidades de cada pareja. No hace falta asumir H-W, sólo asumir que no hay padres fuera de los genotipados. Otro problema es si no te salen padres compatibles y quisieras inferir el genotipo más probable del padre fantasma (y luego buscarlo) o si quieres determinar "a priori" la probabilidad de que unos hijos vengan de una población determinada. Supongo que ahí si que te hace falta asumir que hay equilibrio de H-W o que, si decides que no lo hay, que las frecuencias genotípicas son las que has observado. ¿Qué es lo que hacen programas como CERVUS en estos casos? Otro problema es determinar si la falta de equilibrio es real o se debe a la aparición de alelos nulos, u otro artefacto. Jesús rodrigo.badilla@gmail.com