Fungal infections are a global problem imposing considerable disease burden. One of the unmet needs in addressing these infections is rapid, sensitive diagnostics. A promising molecular diagnostic approach is high-resolution melt analysis (HRM). However, there has been little effort in leveraging HRM data for automated, objective identification of fungal species. The purpose of these studies was to assess the utility of distance methods developed for comparison of time series data to classify HRM curves as a means of fungal species identification. Dynamic time warping (DTW), first introduced in the context of speech recognition to identify temporal distortion of similar sounds, is an elastic distance measure that has been successfully applied to a wide range of time series data. Comparison of HRM curves of the rDNA internal transcribed spacer (ITS) region from 51 strains of 18 fungal species using DTW distances allowed accurate classification and clustering of all 51 strains. The utility of DTW distances for species identification was demonstrated by matching HRM curves from 243 previously identified clinical isolates against a database of curves from standard reference strains. The results revealed a number of prior misclassifications, discriminated species that are not resolved by routine phenotypic tests, and accurately identified all 243 test strains. In addition to DTW, several other distance functions, Edit Distance on Real sequence (EDR) and Shape-based Distance (SBD), showed promise. It is concluded that DTW-based distances provide a useful metric for the automated identification of fungi based on HRM curves of the ITS region and that this provides the foundation for a robust and automatable method applicable to the clinical setting.

No abstract