Current techniques of molecular cloning allow purification of DNA fragments up to 45 kilobases (kb) long in a single clone. With effort, somewhat larger areas can be cloned as overlapping cosmids. There is still a significant gap, however, between gene mapping by these methods and the resolution obtainable by somatic cell genetics and in situ hybridization. These techniques are able to localize a gene to what appears grossly to be a small region of a chromosome, but actually represents 2000-10,000 kb or more. Many interesting questions in medicine, molecular biology, and human genetics hinge upon development of a cloning procedure which is adaptable to this """"""""gap"""""""" region between 50 and 2000 kb. In this application, an approach to this problem with wide potential applicability is proposed. The basic principle is to generate long linear DNA fragments by partial digestion of high molecular weight genomic DNA, and then to ligate these at very low concentrations so that predominantly covalent circles are formed, preferably in the presence of a marker gene. At the junctions DNA fragments which were originally many kilobases apart will be adjacent. The marker gene can then be used to select these junction fragments and generate a phage or cosmid library from them. Screening such a library with a probe of interest will produce another DNA fragment which previously was located many kilobases away in genomic DNA. The feasibility of this approach has already been demonstrated with a model system. Among the immediate applications will be mapping of large gene families such as the major histocompatibility complex of man, crossing chromosome deletions, and moving closer to a gene of interest from a linked gene some distance away, with the eventual goal of cloning the responsible genes themselves in conditions such as Huntington disease or the fragile X syndrome.