DNA barcoding authentication for the wood of eight endangered Dalbergia timber species using machine learning approaches.

Reliable wood identification and proof of the provenance of trees is the first step for combating illegal logging. DNA barcoding belongs to the promising tools in this regard, for which reliable methods and reference libraries are needed. Machine learning approaches (MLAs) are tailored to the necessities of DNA barcoding, which are based on mathematical multivaried analysis. In the present study, eight Dalbergia timber species were investigated in terms of their DNA sequences focusing on four barcodes (ITS2, matK, trnH-psbA and trnL) by means of the MLAs BLOG and WEKA for wood species identification. The data material downloaded from NCBI (288 sequences) and taken from a previous study of the authors (153 DNA sequences) was taken as dataset for calibration. The MLAs' effectivity was verified through identification of non-vouchered wood specimens. The results indicate that the SMO classifier as part of the WEKA approach performed the best (98%~100%) for discriminating the eight Dalbergia timber species. Moreover, the two-locus combination ITS2+trnH-psbA showed the highest success rate. Furthermore, the non-vouchered wood specimens were successfully identified by means of ITS2+trnH-psbA with the SMO classifier. The MLAs are successful in combi- nation with DNA barcode reference libraries for the identification of endangered Dalbergia timber species. [ABSTRACT FROM AUTHOR]