Inactivating mutations in TSC1 and TSC2 cause tuberous sclerosis complex (TSC). This disease is variable in severity and classically causes seizures, intellectual disability, behavioural difficulties, brain tumours, heart tumours, renal tumours and a facial rash. It occurs worldwide and is found in approximately one in 10,000 live births. It can be inherited as an autosomal dominant disorder, but about 70% of cases are new in the family and there is a significant risk of recurrence in a second child because of gonadal mosaicism in one of the parents. In some cases, mildly affected individuals are not recognised till the birth of a severely affected child. Genetic testing is needed both for diagnosis and for reproductive decisions, but in at least 10% of tests the causative change is not found. There are also many TSC variants of uncertain significance which are often unique. The TSC1 and TSC2 databases (www.lovd.nl/TSC1 and www.lovd.nl/TSC2) attempt to record all TSC variants which have been reported, both in the contest of genetic testing for TSC and in other clinical conditions. Classification of the pathogenicity of these TSC variants in the TSC1 and TSC2 databases refers to their ability to cause TSC. Data from August 2015 show 889 different small variants for TSC1 of which 66% were pathological and 9% were not. For TSC2, there were 2522 different small variants of which 50% were pathological and 9% were not. All others were to some extent uncertain. Recent advances in the quantity and quality in NGS, and the access to enormous amounts of population data have produced many challenging opportunities to improve variant classification. We have now formalised our decision-making on the pathogenicity of variants and this uses type and position, likely effect of variant, confidence in diagnosis, frequency of variant in different patients, family details including de novo reports, co-occurrence with known harmful variant, and where appropriate, an in vitro function assay. We now also screen our database automatically for duplicates in very large cohort datasets, most of which do not have extensive phenotype data. We will present our current practice for interpretation and would welcome comments. The TSC variation databases are funded by the TSA and the TS Alliance.