This invention relates to a method and an apparatus for determining a reliability indicator for at least one set of signatures obtained from clinical data collected from a group of samples. The signatures are obtained by detecting characteristics in the clinical data from the group of sample sand each of the signatures generate a first set of stratification values that stratify the group of samples. At least one additional and parallel stratification source to the signatures obtained from group of sample sis provided, the at least one additional and parallel stratification source to the signatures being independent from the signatures and generates a second set of stratification values. A comparison is done for each respective sample, where the first stratification values are compared with a true reference stratification values, and where the second stratification values are compared with the true reference stratification values. The signatures are assigned with similarity measure indicators indicating whether the first and the second stratification values match with the true reference stratification values. These are then implementing as input in determining the reliability of the signatures.