Relationship coefficients are traditionally based on pedigree data. Today, with the development of molecular techniques, they are often completely replaced by coefficients calculated from molecular data. Examples are relationships from microsatellites for biodiversity studies but also genomic relationships from SNP as currently used in genomic prediction of breeding values. There are, however, many situations in which optimal combination of both sources would be the best solutions. Obviously, this is the case for incompletely genotyped populations, but also when pedigree information is sparse. Also, markers, even dense ones, do not reflect the whole genome and therefore give only an incomplete picture of relationships. The main objective of this study was therefore to develop a method to calculate a relationship matrix by the combination of molecular and pedigree data. It will be useful for all situations where pedigree and molecular data are available. In this study, based on simulations of pedigree and marker data, we used partial least squares regression and linear regression to combine total allelic relationship coefficients calculated for each marker with additive relationship coefficients calculated from incomplete pedigree. The results showed that the greatest advantage of this method, compared with the one that replaces a part of the pedigree-based relationship matrix by a genomic relationship matrix, is that adding the partial pedigree data allows for the correction of the molecular coefficient for the ungenotyped part of the genome.