By hybridizing the functional-domain and sequence-correlated pseudo amino acid composition approaches, a 2-layer predictor called "Quat-2L" was developed for predicting the quaternary structural attribute of a protein according to its sequence information alone. The 1st layer is to identify the query protein as monomer, homo-oligomer, or hetero-oligomer. If the result thus obtained turns out to be homo-oligomer or hetero-oligomer, then the prediction will be automatically continued to further identify it belonging to one of the following six subtypes: (1) dimer, (2) trimer, (3) tetramer, (4) pentamer, (5) hexamer, and (6) octamer. The overall success rate of Quat-2L for the 1st layer identification was 71.14%; while the overall success rates of the 2nd layer for homo-oligomers and hetero-oligomers were 76.91 and 82.52%, respectively. These rates were derived by the jackknife cross-validation tests on the stringent benchmark data set in which none of proteins has ≥ 60% pairwise sequence identity to any other in the same subset. As a web-server, Quat-2L is freely accessible to the public via http://icpr.jci.jx.cn/bioinfo/Quat-2L, where one can get 2-level results in about 15 s.