ellmen / alcov Goto Github PK

Dear,

I tried the installation both suggested ways but I get error:
ImportError: cannot import name 'pywrapinit' from 'ortools.init'

I was thus wondering if this needs to be adapted in the code? Or if I need a particular python/pip version to use your tool?

Thank you

Dear Isaac,
Thank you for sharing your your solution to predict Pango lineage composition in a pooled wastewater sample.
Here are a couple of questions and suggestions.

From the code we see that you fit a linear model between selected mutations abundance in the inputs (Y) and mutation profiles defining selected Pango Lineages (X).

1. Why only 12 lineages were selected for querying (B.1.526.1, B.1.427, B.1.617.3, B.1.617.2, B.1.526, B.1.429, B.1.617, B.1.1.7, P.1, B.1.617.1, B.1.351, B.1.526.2, P.2)? Could you provide logic on the mutations and lineages selection process? Why these were selected and not all ~1200+ Pango Lineages?
2. It seems that only lineages with S:N501Y mutation were selected, was there a reason for this? Why this filter is necessary?
3. How Pango lineage mutation profiles were defined? I see that you've used Outbreak.Info API queries by lineage such as https://api.outbreak.info/genomics/lineage-mutations?pangolin_lineage=B.1.1.7 to pull data. What that the case?
4. Does the future iteration of the model would consider multiple testing corrections and intrinsic hierarchical structure existing between lineages given rise to increased error rates and inflated beta coefficients?
5. The beta coefficients of Panglo lineage terms are presented as abundances, but are just weights of the model terms. Do you plan also to assign p-values to terms and do some randomization (e.g. permutations) to improve reliability and stability of the prediction?
6. Do you also plan to update setup.py to automatically setup dependencies numpy, scikit-learn, matplotlib seaborn and pysam during install?
7. What if a given query position has more than one mutation? Are you only taking into account abundance of the lineage-associated SNV while disregarding other potential mutations?
8. When selecting mutations "strongly" associated with Pango lineages mut_lins[aa_m][l] > 0.5, why global mutation prevalence of 0.5 was selected as a cutoff and not more higher values such as 0.9 in case you would like to look for "lineage semi-exclusive" mutations? Would you consider adding this as a command-line parameter?
9. Have you considered training a tree ensemble classifier (XGBoost, RF) or conditional inference forest (CIF) on the filtered GISAID input?

Good luck with this project and we would be glad to exchange ideas and test other approaches on lineages resolution from the mixed wastewater samples.

Kirill

Hello,

I am having an issue where lineages added to the lineages.txt file are not accepted by alcov, for example XBB.1.22 I get the following error:

File "/usr/local/lib/python3.8/dist-packages/alcov/lineages.py", line 375, in

lin_mut_profiles = [[round(mut_lins[mut][lin]) for mut in covered_muts] for lin in lineages]

KeyError: 'Omicron (XBB.1.22-like)'

Thank you.