Probabilistic peak detection in CE‐LIF for STR DNA typing

Abstract

In this work, we present a novel probabilistic peak detection algorithm based on a Bayesian framework for forensic DNA analysis. The proposed method aims at an exhaustive use of raw electropherogram data from a laser-induced fluorescence multi-CE system. As the raw data are informative up to a single data point, the conventional threshold-based approaches discard relevant forensic information early in the data analysis pipeline. Our proposed method assigns a posterior probability reflecting the data point's relevance with respect to peak detection criteria. Peaks of low intensity generated from a truly existing allele can thus constitute evidential value instead of fully discarding them and contemplating a potential allele drop-out. This way of working utilizes the information available within each individual data point and thus avoids making early (binary) decisions on the data analysis that can lead to error propagation. The proposed method was tested and compared to the application of a set threshold as is current practice in forensic STR DNA profiling. The new method was found to yield a significant improvement in the number of alleles identified, regardless of the peak heights and deviation from Gaussian shape.

An artificial neural network system to identify alleles in reference electropherograms

“…In order to compare results to other reading software programs a system of peak centre identification and classification was required. We employ the method of [4] whereby a likelihood ratio is assigned to each scan point:…”

“…where H1 is that there is a peak centre at the scan point, H2 is that there is not a peak centre at the scan point (either because there is no peak nearby, or a peak nearby but with a centre at a different scan point) and D is a window of data that acts as context (in the case of [4] and our work here the window was 21). We made the following adjustment to the algorithm:  A peak centre was designated as any scan where the arbitrary value 10…”

“… The amplitude of the peaks ('A' from [4]) was designated as the observed height (after baselining) of the scan being examined i.e the central scan for H1 or the scan being considered a peak centre when not central in H2…”

“…We have demonstrated the performance of an ANN system, coupled with the Bayesian peak detection algorithm of [4] on EPGs that range from very weak to overloaded. Profiles 1 and 5…”

“…A recent work by Taylor et al [1] demonstrated an artificial neural network (ANN) that was trained on two good quality reference EPGs to classify data in the 6-FAM dye lane and then applied to a third (also good quality) EPG with reasonable success. Taylor et al [1] provided a proof of concept that ANN could be used to interpret EPGs, which we extend here by: 1) Increasing the amount of training data 2) Increasing the range of training EPG quality from completely blank to highly overloaded 3) Improving on the architecture of the ANN used 4) Training a series of ANN that are used on different areas of the EPG 5) Coupling the predictions of the ANNs with a peak detection algorithm originally designed for LCMS data [2,3] and recently extended to DNA EPG data [4] to produce a peak detection and classification system Having created the peak detection and classification system we trial it on several profiles and demonstrate the results, which we compare to the peaks flagged by Genemapper® ID-X (Life Technologies).…”