Аннотация:In this paper research in the field of application multiprocessor systems for genome assemblies reconciliation has been carried out. A large number of algorithmic approaches aimed to solve the task of de novo assembly from short reads, however the results of their work on the same raw data often differ essentially. A parallel algorithm for merging two or more assemblies without relying on a reference genome is presented. Due to the large data volume the computations in the distributed memory model on computational cluster are required. The proposed method integrates a combination of draft assemblies reducing resulting contigs fragmentation. Sequential version of the algorithm is implemented in C/C++ and is available at https://bitbucket.org/kromanenkov/gar.