The question of HLA susceptibility to Type 1 (insulin-dependent) diabetes mellitus remains unresolved. In the present study, 127 diabetic patients and 177 unrelated control subjects have been analysed for their class I and class II serological antigens, class II (DR, DQ) DNA restriction fragment length polymorphisms and DQA1 and B1 exon-2 nucleotide sequences and their corresponding amino acid residues. By using the aetiologic fraction (delta) as an almost absolute measure of the strongest linkage disequilibrium of an HLA marker to the putative Type 1 diabetes susceptibility locus, it has been found that the strength of association of the HLA markers may be quantified as follows: DR4 less than DR3 less than DR3 or DR4 less than non-Aspartate 57 beta DQ and Arginine 52 alpha DQ less than Arginine 52 alpha DQ. Thus, molecular HLA-DQ markers appear to be more accurate as susceptibility markers than the classic serologically defined ones (DR3 and DR4); however, any effect of DQ markers disappears when non-DR3/DR4 individuals are considered, suggesting that DR factors (or others in between DQ and DR) are also important. In addition, a dominant non-Aspartate 57 beta DQ susceptibility theory does not hold (but a recessive one does) in our diabetic population (probably due to the high frequency of the protective DR7-non-Aspartate 57 beta DQ haplotypes); Arginine 52 alpha DQ is the best single HLA marker found in our population, both as a recessive or as a dominant one. Also there are 13 patients in our sample who bear neither Arginine 52 alpha DQ nor non-Aspartate 57 beta DQ susceptibility factors.(ABSTRACT TRUNCATED AT 250 WORDS)