We present a model fitting the distribution of non-systematic errors in STR second generation sequencing, SGS, analysis. The model fits the distribution of non-systematic errors, i.e. the noise, using a one-inflated, zero-truncated, negative binomial model. The model is a two component model. The first component models the excess of singleton reads, while the second component models the remainder of the errors according to a truncated negative binomial distribution.

We estimated the parameters of the model in two ways: (1) we maximised the likelihood using an explicitly calculated gradient function and (2) we used the expectation-maximisation, EM, algorithm. The estimated parameters were used to create dynamic, sample specific thresholds for noise removal using marker specific proportions of the negative binomial distribution.

Based on data from dilution series experiments (amounts of DNA ranging from 100 pg to 2 ng) conducted at The Section of Forensic Genetics, Department of Forensic Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Denmark, the method was compared to that of a naïve model that implies the removal of reads with a coverage of less than 5–10% of the total marker coverage. In comparison, our method resulted in three allelic drop-outs (true alleles below threshold), whereas the 10%-threshold induced 12 drop-outs. The non-filtered error reads (e.g. stutters, shoulders and reads with miscalled bases) will subsequently be modelled by different statistical methodologies.