Error Analysis of Deep Sequencing of Phage Libraries: Peptides Censored in Sequencing
(a) Operator and multiset description of the error filtering procedure. Applying a Phred > 30 cutoff to library filtered by Phred>1 cutoff () yields a subpopulation of the library (). If errors are sequence-independent, the process should be identical to random sampling (). Any sequence-specific bias () should be detected as deviation from . (b) Progressive sampling with more stringent cutoff. (c) Theoretical and theoretical 99.9% confidence interval (blue). (d) Observation of statistically significant deviation from operator: dots beyond the blue line represent sequences prone to bias. Red dots represent sequences that disappeared after in process or during sampling. (e) Magnitude of the bias range from 5 to 100-fold. (f) Bias in sampling of Phred > 30 data from Phred > 1 data ((f) is theory, (g) is observed). (h) Bias upon sampling of Phred > 30 data from Phred > 13. Many sequences were lost in this sampling and this loss was statistically significant beyond the 99.9% interval. This result shows that some sequences have propensity to harbor low- and medium-quality reads. Distribution of the errors is sequence specific.
Article of the Year Award: Outstanding research contributions of 2020, as selected by our Chief Editors. Read the winning articles.