Karyotype frequencies

[alimanfoo]

For Ag3 we now have a karyotype() function which infers a karyotype for each sample at each of several possible inversions on Chromosome 2. Suggest to also add a karyotype_frequencies() and karyotype_frequencies_advanced() function which compute frequencies within cohorts, analogous to existing functions for SNPs and CNVs.

[sanjaynagi]

Havent got time to implement but heres some basic code i had.

def calculate_karyo_frequencies(inversion, sample_sets, sample_query, cohort_column='cohort_admin2_year'):

    df_karyo = ag3.karyotype(inversion, sample_sets=sample_sets, sample_query=sample_query)
    df_samples = ag3.sample_metadata(sample_sets=sample_sets, sample_query=sample_query)
    df_karyo = df_karyo.merge(df_samples, how='left', on='sample_id')

    afs = {coh:calc_allele_freq(df_karyo.query(f"{cohort_column} == @coh")[f'karyotype_{inversion}']) for coh in df_karyo[cohort_column].unique()}
    counts = {coh:df_karyo.query(f"{cohort_column} == @coh").shape[0] for coh in df_karyo[cohort_column].unique()}
    af_df = pd.DataFrame(afs).T.reset_index().rename(columns={'index':cohort_column, 0:inversion, 1:'alt_alleles', 2:'total_alleles'})
    af_df[f'{inversion}_standard'] = 1 - af_df[inversion]
    df_af = af_df.merge(df_samples[['latitude', 'longitude', 'year' , 'taxon', cohort_column]].drop_duplicates(cohort_column), how='left', on=cohort_column)

    return df_af, counts

[jonbrenas]

I think, similarly to what we have with other frequency functions, we will have 2 functions karyotype_frequencies (that takes a cohort parameter) and karyotype_frequencies_advanced (that takes a area_by parameter and a period_by parameter). In both cases, there would be 3 rows/variants (hom. ref., het., hom. alt.) and the columns would be the frequencies (+ max_af, probably), the contig and name of the inversion. Any other important data that needs to be added?

I think it would make sense to have multiple inversions in the same heatmap in the frequency functions (the question then is do we modify karyotype to take a list of inversions or do we loop the function in the `karyotype_frequencies[_advanced]" functions)

The frequency functions is going to go in the same folder, if not file, as the karyotype function, in my opinion.

[alimanfoo]

Just capturing some discussion here, maybe better to call this "inversion_frequencies()" or "inversion_allele_frequencies()" to be clear it is about computing the frequency of the inverted allele within a cohort.

[jonbrenas]

The first prototype for the first function yields something that looks like this:

[31puneet]

Hi @jonbrenas, I see that PR #732 is currently in draft. Is this still being actively worked on? I'd be interested in contributing to this issue if help is needed.

[jonbrenas]

I don't really remember where we stopped on this one. You are welcome to take a look and make the changes that you think are appropriate.