Very rough analyses testing epitope deconvolution for libA and libB antibody cocktail + AUSAB-13 selections. Mostly playing around with different similarity regularization weights.
Note one change to the analysis. I had temporarily dropped the minimum counts + fraction for variants in the no-antibody control, so that the cutoff was less stringent for prob_escape calculations. This was initially done to accommodate low sequencing coverage, and then minimum counts was increased back to 15. But I neglected to also raise the minimum fraction. So I re-analyzed data with this fraction raised 100-fold to 2e-6 (which equals 0.1 / 5e4, the rough number of functional variants in the larger library).
Very rough analyses testing epitope deconvolution for libA and libB antibody cocktail + AUSAB-13 selections. Mostly playing around with different similarity regularization weights.
Note one change to the analysis. I had temporarily dropped the minimum counts + fraction for variants in the no-antibody control, so that the cutoff was less stringent for prob_escape calculations. This was initially done to accommodate low sequencing coverage, and then minimum counts was increased back to 15. But I neglected to also raise the minimum fraction. So I re-analyzed data with this fraction raised 100-fold to 2e-6 (which equals 0.1 / 5e4, the rough number of functional variants in the larger library).