Classification of cryptic Lepidoptera species through near and middle infrared spectroscopy and multivariate analysis techniques

Cryptic species are defined as two or more distinct species that are erroneously classified and hidden under one species name. They represent a taxonomic challenge for biologist and although the advent of relatively less expensive DNA sequencing techniques pose a new tool for differentiating morphologically similar species , these require removal and consequent loss of body appendages of a sometimes rare and scarce specimen. In this matter, vibrational spectroscopy techniques, as near and middle infrared spectroscopy, have the inherent advantage of being fast, low cost, non-invasive and non-destructive, which is highly suitable for the study of delicate and valuable samples.
C. cumbre (Schaus, 1902) and C. triunviralis (Hayward, 1939) are two cryptic species of Lepidoptera of the genus Cumbre Evans, 1955 genus (Hesperiidae: Hesperiinea: Moncini). Morphologically they can scarcely be separated by features of masculine and feminine genitalia. Both species inhabit the south and south east of Brazil, C. triunviralis also inhabits Paraguay, Argentina and Uruguay. In the present work we tested the capacity of infrared spectroscopy technique to identify these cryptic species. Two hundred adult specimens, fifty males and fifty females of each species were obtained from several geographic origins. The insects were collected between 1940 and 2013, identified and maintained in the Zoology Department of the Federal University of Paraná; therefore they represent a pool of samples of broad temporal and spatial variability. The measurements were carried out with an Excalibur Bio-Rad FTS 3500 GX series spectrophotometer (Bio-Rad laboratory, Cambridge, MA, USA) equipped with a KBr light source and a DTGS detector, operating between 400 and 7500 cm-1 wavenumber, with a spectral resolution of 1 cm-1 and 32 scans. Samples were positioned with the point of incidence of the beam on the rear right wing, at the end of the discal cell. Multivariate analysis of spectroscopic data was carried out using the software Pirouette (Infometrix, USA). Partial least squares–discriminant analysis (PLS-DA) and soft independent modeling of class analogy (SIMCA) methods were used to classify both species. fourty samples (10 females and 10 males of each species) were separated to create an external validation set, leaving the remaining 160 samples as the calibration set. The sets were also used to create a model for gender classification, using SIMCA. Spectra were transformed using mean-center, and preprocessed through multiplicative scatter correction (MSC) or smoothing and first derivative.
Both models, PLS-DA and SIMCA were successful in the classification of the external validation set (100 % of samples correctly classified in both cases), with no outliers found, which considering the high temporal and spatial variability of the collected samples, show the models are robust. The following frequencies 410-418 cm-1, 1490-1498 cm-1 ; y 7230-7240 cm -1, were related to the highest coefficients in PLS-DA regression vector and were also the variables with the highest discriminant power in the SIMCA model, therefore these regions could be related to molecular differences between both species, and could be used as discriminatory features between the cryptic pair.