Genotypes in probabilities?

[Hoeze]

Hi, how do I retrieve the genotype column ID for genotype[x].compute()["probs"]?

[horta]

Hi @Hoeze , you can use x to index bgen["variants"] and retrieve the information you want about the x-th variant: bgen["variants"].iloc[x].

[Hoeze]

Thanks for your fast response @horta.
I tried your suggestion like in the following example:

In [22]: from bgen_reader import *                                                                                                                                                                                                          

In [23]: bgen = read_bgen("complex.bgen")                                                                                                                                                                                                   
Reading samples|===========================================================================================================================================================================================================================|
Mapping variants: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 10/10 [00:00<00:00, 51590.46it/s]

In [24]: geno = bgen["genotype"][8].compute()                                                                                                                                                                                               

In [25]: v = bgen["variants"].compute().iloc[8]                                                                                                                                                                                             

In [26]: print(v)                                                                                                                                                                                                                           
id                                                
rsid                                            M9
chrom                                           01
pos                                              9
nalleles                                         8
allele_ids    A,G,GT,GTT,GTTT,GTTTT,GTTTTT,GTTTTTT
vaddr                                          783
Name: 8, dtype: object

In [27]: print(geno["probs"])                                                                                                                                                                                                               
[[ 1.  0.  0.  0.  0.  0.  0.  0. nan nan nan nan nan nan nan nan nan nan
  nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan]
 [ 0.  0.  0.  0.  0.  0.  1.  0. nan nan nan nan nan nan nan nan nan nan
  nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan]
 [ 0.  0.  0.  0.  1.  0.  0.  0. nan nan nan nan nan nan nan nan nan nan
  nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan]
 [ 0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  1.  0.
   0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.]]

In [28]: print(geno["probs"].shape)                                                                                                                                                                                                         
(4, 36)

I already found the allele_ids, but how do I know which column in geno["probs"] contains which genotype?
For example, column 1 could contain A/A, A/G, etc.

In the end, my target is to extract in a certain genomic range the most likely sequence for each individual.

[horta]

That is the tricky part because BGEN allows for very general genotype. It depends wether you have Unphased genotype, Phased genotypes, number of alleles, ploiyd. The quick start gives a quick idea on how to perform the association between probability and genotype (in particular, read the comments in the code). But the ultimate answer is in the section Per-sample order of stored probabilities of bgen specification.

[Hoeze]

Hm, I see. How do you solve this problem in your projects?
Do you have a method which calculates it?
Otherwise, I could try to write one...

[horta]

Have a look at allele_expectation: https://bgen-reader.readthedocs.io/en/latest/expectation.html

[horta]

But make sure you have unphased genotype: "This function supports unphased genotypes only."

[Hoeze]

Ah, thank you for the hint, I found what I searched for:

bgen-reader-py/bgen_reader/_helper.py

Line 1 in 2651d87

def get_genotypes(ploidy, nalleles):

Would it be possible to have this function publicly exported in genotype?

[horta]

Sure. Will do it for the 3.1.0 release