Bioinformatics pipelines designed for paired analysis of tumour and matched normal samples use germline variants detected from normal tissue samples to identify bona fide tumour somatic mutations. Unexpected contamination of normal samples with tumour cells can lead to the subtraction of genuine somatic variants, reducing variant detection sensitivity and compromising downstream analyses. Leveraging the power of whole-genome sequencing data available from the 100,000 Genomes Project at Genomics England, we developed the Tumour in Normal Contamination (TINC) tool for TIN contamination assessment. TINC was validated in silico using patient-derived whole-genome samples (R2=0.95; p<2.210-16 and R2=0.85; p<3.310-13 for expected versus observed contamination in two tumour types), and against orthogonal minimal residual disease (MRD) testing with 70 leukaemia samples. From a systematic review of whole-genome data available for 771 patients with haematological malignancies and sarcomas, we found TIN contamination (>1%) across a range of cancer clinical indications and DNA sources. The highest prevalence was among saliva samples from acute myeloid leukaemia (AML) patients (43/114 samples, 38%), and sorted CD3+ T cells from myeloproliferative neoplasms (MPN) patients (22/24 samples, 91%). Further exploration of genomic data for AML and MPN patients has revealed 108 hotspot mutations in genes with high prevalence of somatic alterations in haematological cancers (JAK2, FLT3, DNMT3A, TP53, KIT, NRAS and IDH2), 27 of which supported by at least 5% of reads in the normal sample and therefore at risk of being subtracted by standard pipeline for somatic variant calling. Our work highlights the importance of contamination assessment for accurate detection of somatic variants in both research and clinical settings. TINC allows accurate and highly sensitive quantification of contamination levels, giving confidence in the validity of the whole genome assessment of the tumour. We propose this to become an essential metric in clinic practice alongside other routine quality control measures, especially as large-scale pan-cancer sequencing projects are adopted to deliver accurate data from which to make clinical decisions for patient care.