BUSCADOR RECOLECTA

Digital.CSIC. Repositorio Institucional del CSIC

oai:digital.csic.es:10261/330863

Supplementary Material 7: Table of bivariate correlations of the first 15 variables chosen by the Random Forest algorithm according to their Mean Decrease Accuracy values (See Supplementary Material 6). Number in brackets indicate selection order. Red values have Person correlation coefficient >0.50| and p-value <0.05, and considered to be correlated variables. Number in brackets relate to the variable selection order for subsequent analyses., Crop wild relatives are species related to cultivated plants, whose populations have evolved in natural conditions and confer them valuable adaptive genetic diversity, that can be used in introgression breeding programs. Targeting four wild lentil taxa in Europe, we applied the predictive characterization approach through the filtering method to identify populations potentially tolerant to drought, salinity, and waterlogging. In parallel, the calibration method was applied to select wild populations potentially resistant to lentil rust and broomrape, using, respectively, 351 and 204 accessions evaluated for these diseases. An ecogeographic land characterization map was used to incorporate potential genetic diversity of adaptive value. We identified 13, 1, 21, and 30 populations potentially tolerant to drought, soil salinity, waterlogging, or resistance to rust, respectively. The models targeting broomrape resistance did not adjust well and thus, we were not able to select any population regarding this trait. The systematic use of predictive characterization techniques may boost the efficiency of introgression breeding programs by increasing the chances of collecting the most appropriate populations for the desired traits. However, these populations must still be experimentally tested to confirm the predictions., Peer reviewed