[A probe for evaluating the role of HLA class II antigens coded with the DRB gene in etiopathogenesis of familial sarcoidosis in Poland using the K-nearest neighbor experimental statistical method].

The aim of this study was evaluation of the significance of HLA-DRB1,-DRB3,-DRB4,-DRB5 associated alleles in the genetic susceptibility to sarcoidosis. We investigated 17 Polish 'families with' familial occurrence of the disease. Thirty three affected family members and 78 healthy relatives and 101 healthy individuals (control group) have been typed for HLA class II DR antigens. Assuming that relatives from families with two or more affected members are more susceptible to develop sarcoidosis we considered two classes: affected and healthy family members taken together (class 1, N = 111) and healthy control group (class 2, N = 101). HLA antigens from both classes were compared using a statistical pattern recognition method. This method, called k-NN method, assumes that objects (individuals) are described by a certain number of variables called features. Selected features that played an important role in the decision to which class an individual person would be classified were: HLA-DR7, HLA-DR2(16), HLA-DR1, HLA-DR5(12), HLA-DR6(14), HLA-DR6(13), HLA-DR9, HLA-DR5(11) and HLA-DRw52. K-NN method allowed to classify properly 76% of studied subjects to healthy or disease susceptible group. However, 24 out of 100 individuals would be misclassified which gives the total error rate of 0.24. We concluded that using HLA-DR antigens as features characterising every individual we can predict with high probability to which class ("high risk" or "healthy") individual would belong.