search

esrud / currentNe

Estimation of current effective population using artificial neural networks
GNU General Public License v3.0
3 stars 0 forks source link

k not estimated, F too large #1

Open melisaolave opened 9 months ago

melisaolave commented 9 months ago

Hi, thanks for making this available to the community!

I tried running currentNe in my dataset with 22 chromosomes, 1,517,015 SNPs, 16 individuals. I wanted the program to estimate the full siblings from the data, but it returned: Number of full siblings that a random individuals has in the sample: Not estimated because F value is too large

and returned this about F: Estimated F value of the population (deviation from H-W proportions): -0.23442383

Estimated pop size is far to small, with Ne = 18. It makes no sense.

F value (inbreeding coefficient, right?) might be high, but still seems normal to me, and it is comprable to other closely related species. Why is not possible to estimate the number of full siblings? is there anything I can do?

thanks! Melisa

esrud commented 9 months ago

There are many more heterozygotes in your sample than expected under the assumption of a single random mating population. In this situation, the sibling identification method implemented in currentNe is not reliable. However, the k-value can be estimated using other published methods (e.g. Jones & Wang 2010 MER).

Best, Enrique

melisaolave commented 9 months ago

Hi Enrique,

thanks for your response. So, I suppose that the estimate of Ne is also not reliable wiht currentNe?

all the best,

Melisa

esrud commented 9 months ago

There are several factors that affect Ne estimates (see figures S2 and S3 in the supplementary material of the currentNe paper). If you think they are not involved in your case, then the Ne estimate might be fine, even though it assumes k=0 unless you provide a different value of k obtained by some other method.

Best, Enrique