matthiaszeller
commented
5 years ago Clinical data
- I put the same code for
Mixed genotype detectedandMixed - I kept only the variables you specified
Viral data
I plotted the missing rates per individuals and per variant. I dropped the variants that had a missing rate over 15% which removes 168 variants (out of 5393). I didn't drop any individual since the worst case has about 15% of missing values.
General remarks
I'm using Python version 3.7.3
Viral data.ipynb
You are looking at NA values. Plot 54 is great. It shows that the NA's do not occur randomly distributed over the genome, but at specific locations.
Additionally, you could check the missing frequency (or non missing frequency) by individuals. You do the mutation rate calculation for VARIANTS + INDIVIDUALS, just do the same with NA's. This is useful because we want to make sure to exclude individuals that have too many NA's (an indication for bad quality sequencing).
Clinical data.ipynb
df['GT'].value_counts()there are two levels that are calledMixed genotype detectedandMixed. This seems like they should be coded the same. Can you use a common level for any downstream analysis?IGM_ID(host id),gilead_id(pathogen_id),GT,COUNTRY,ETHNICITY,RACE,SEX,AGE,OAV_EXPERIENCE,BASELINE_HBVDNA_IU/mL,BASELINE_HBVDNA_Dil_IU/mL,BASELINE_HBEAG_STATUS.Plink introduction.ipynb
You can start exploring the association analysis done in PLINK: https://www.cog-genomics.org/plink/1.9/assoc.
They have less and more sophisticated models. We will start with a simple model (one outcome, one predictor:
y ~ x) and then extend this model with covariates.You will also need to do QC on the genotype data side. Keywords here are missing genotypes, missing individuals, minor allele frequency, Hardy Weinberg equilibrium.
General remarks
pandas-plink.