Supervised machine learning and genotype by trait biplot as promising approaches for selection of phytochemically enriched Rhus coriaria genotypes

Sumac is considered as a medicinal and industrial plant. Climate change threats natural ecosystems and hence, evaluation of sumac's genetic diversity, identification of superior genotypes, and conservation of such materials is important. In this study, 5 wild populations of sumac were investigated. Fruits of 75 sumac genotypes (15 genotype per population) were analyzed using HPLC-LC/MS-MS method. Likewise, genomic DNA of 75 genotypes were fingerprinted using 18 ISSR primers. Analysis of variance revealed significant genetic variability among studied populations of sumac considering malic acid, malic acid hexoside 2.71, malic acid hexoside 6.11, coumaric acid, ellagic acid11.49. Malic acid was identified as phytochemical marker in sumac fruit which can be implemented for screening sumac genotypes even from the same population. Genotype by trait analysis revealed V6, V10, D10, D14, A1, A14, K3, K15, N10, and N11 as top-performing genotypes (winners) which possessed the majority of phytochemical constituents in highest value. Here, the identified phytochemically superior sumac group was effectively distinguished from the inferior sumac group using ISSRs information via supervised machine learning. By using 13 feature selection algorithms, ISSR loci (U823) L1, (U835) L1, (U801) L1, (U816) L2, (U816) L4, (U835) L4, (U854) L1, and (U835) L9 were identified as functional markers which could predict phytochemical response of sumac germplasm. In conclusion, there is vast range of phytochemically divergent sumac genotypes in its natural habitats that could effectively recognized in any season by merging artificial intelligence with genomic information.