High-throughput interrogation of programmed ribosomal frameshifting in human cells

Martin Mikl; Yitzhak Pilpel; Eran Segal

Second, for bins with a read count of less than three, the bin value was set to zero. Third, for each 861 variant we set to zero bins surrounded by zero values as these constituted isolated bins unlikely to 862 come from the actual distribution. Forth, to reduce bias coming from the "open end"-bins at the 863 extreme values, we set their count to match their neighbor's if it was higher, as these bins are defined 864 as containing the tails of the distribution of variants with no GFP fluorescence and maximal GFP 865 expression, respectively, and peaks in these extreme bins are considered experimental noise. For 866 each variant, we normalized the values across the 16 bins by dividing the filtered read counts for each 867 bin by the sum of all filtered read counts. To determine whether a distribution is unimodal we applied 868 a Savitzky-Golay filter for smoothing the data and removing local maxima introduced by experimental 869 noise. We detected peaks in the smoothed distribution by a simple approach in which a point is 870 considered a maximum peak if it has the maximal value, and was preceded by a value lower by 5%. 871