Cancer scientists overestimate the extent to which high-profile preclinical studies can be successfully replicated, new research from McGill University suggests.