ABSTRACT - Based on its nutritional and economic value, the lima bean (Phaseolus lunatus L.) is the second most important species of the genus. It has high genetic diversity and potential for production and is considered an alternative food and income source. The objective of this study was to apply the restricted maximum likelihood/ best linear unbiased prediction (REML/ BLUP) methodology to estimate genetic parameters and predict genotypic values in F3 populations of lima beans. Twelve characteristics were evaluated in five populations with indeterminate growth habits (H39, H72, H53, H90, and H56). Model 83 from the Selegen program was used for analysis. Considering the genetic parameters, the highest values of genetic variance were for plant height and number of pods per plant. Pod thickness and seed width are favorable for breeding programs. Seed width selection gain was significant for populations H56 and H90 at 11.26 mm and 10.50 mm, respectively. As for the length and thickness of seeds, the gains were less significant, with population H53 showing the greatest gain. The REML/ BLUP methodology proved efficient in estimating genetic parameters and predicting gains in lima bean populations. The estimated selection gains indicated that the highest gains were obtained for plant height, the number of pods per plant, pod thickness, seed width, and the number of days to maturity. Populations H53 and H56 stood out for having large and white seeds, thus being potential populations for species improvement.

ABSTRACT Loss of genetic variability in natural populations affects negatively the adaptive and evolutive process of plant species. In this context, the objective of this work was to evaluate the existing genetic variability between and within natural populations of P. platycephala, using biometric characters of fruits and seeds and physiological characters of seeds and seedlings, and, thus, assess the potential for selection of mother trees, indicate areas for seed collection and use, and generate support for conservation and pre-breeding of plant species. A completely randomized experimental design was used, with 45 treatments (progenies) and four replications of 25 seeds. The progenies were evaluated in laboratory, using seeds collected from 45 mother trees sampled in three natural populations. The following characters were evaluated: seed germination percentage, germination speed index, and mean germination time; shoot and main root lengths; and seedling shoot and root dry weights. The data were subjected to RELM/BLUP analysis for estimating genetic parameters and correlations and to analyses of genetic divergence (UPGMA). The populations present genetic differences to each other and a significant variation between progenies for all characters evaluated, and the variation was greater within the populations. The characters evaluated present high genetic control; therefore, they can be used in analysis of genetic variability of populations and progenies of P. platycephala, thus assisting in the identification of areas for seed collection, conservation programs, and pre-breeding of the species.