Iran is a center of origin and diversity for walnuts (Juglans regia L.) with very good potential for breeding purposes. The rich germplasm available, creates an opportunity for study and selection of the diverse walnut genotypes. In this study, the population structure of 104 Persian walnut accessions was assessed using AFLP markers in combination with phenotypic variability of 17 and 18 qualitative and quantitative traits respetively. The primers E-TG/M-CAG, with high values of number of polymorphic bands, polymorphic information content, marker index and Shannon's diversity index, were the most effective in detecting genetic variation within the walnut germplasm. Multivariate analysis of variance indicated 93.98% of the genetic variability was between individuals, while 6.32% of variation was among populations. A relatively new technique, an advanced maximization strategy with a heuristic approach, was deployed to develop the core collection. Initially, three independent core collections (CC1-CC3) were created using phenotypic data and molecular markers. The three core collections (CC1-CC3) were then merged to generate a composite core collection (CC4). The mean difference percentage, variance difference percentage, variable rate of coefficient of variance percentage, coincidence rate of range percentage, Shannon's diversity index, and Nei's gene diversity were employed for comparative analysis. The CC4 with 46 accessions represented the complete range of phenotypic and genetic variability. This study is the first report describing development of a core collection in walnut using molecular marker data in combination with phenotypic values. The construction of core collection could facilitate the work for identification of genetic determinants of trait variability and aid effective utilization of diversity caused by outcrossing, in walnut breeding programs.