Understanding the cause and effect relationships between mutation and phenotypic profiles of tumour cohorts is hampered in datasets that encompass a diversity of tumour phenotypes and driver mutations. Here we present a novel permutation based strategy for dissecting these relationships using murine leukaemia virus (MuLV) driven lymphoma as a model system. In mice infected with MuLV, the virus replicates throughout the hematopoietic compartment and proviral integrations deregulate nearby genes, giving rise to lymphoid malignancies with 100% penetrance. Because virus/genome junctions can be specifically amplified using ligation mediated PCR, this allow the mutation profile of hundreds of tumours to be determined with minimal coverage at unparalleled sensitivity. From a panel of 500 BCL2 transgenic mice we have identified nearly 3,000 clonal mutations and an additional 1,800,000 subclonal mutations over a spectrum of B and T lymphoid malignancies. The set of mutations are shuffled between tumour samples whilst constraining for both mutation number and clonality in each sample. The output of these permutations are joint distributions of co-mutation and phenotype-genotype interactions that avoid false positives created when analysing frequently and rarely mutated loci together, and is more conservative than traditional contingency table test approaches. Shuffling constraints can also be placed on subtle tumour phenotype differences such that larger cohorts can be analysed in concert without creating false associations due to unrecognised heterogeneity.