Dear Sir,We read with great interest the article by Zeng et al. 1 Basing their study on predictions of hydrophilicity, surface accessibility and buried residues, they restricted their search for NY-ESO-1 B-cell epitopes to the first N-terminal 74 amino acids and identified a dominant B-cell epitope that resides in the first 40 amino acids of the NY-ESO-1 molecule, with amino acids 11-30 representing the core of this B-cell epitope. We also analyzed the epitope specificity of anti-NY-ESO-1 antibodies in 12 sera from cancer patients (12 melanoma, 1 non small cell lung cancer, 1 breast cancer) using recombinant protein fragments of NY- and matrix-bound synthesized decamer peptides with 5 amino acid overlaps between every 2 consecutive decamer peptides covering the entire NY-ESO-1 molecule. Our results confirm the finding that most sera react with the epitope described by Zeng et al. 1 However, our scanning of the entire NY-ESO-1 molecule irrespective of predicted hydrophilicity, surface accessibility and buried residues revealed that there is a second epitope against which IgG antibodies in 7 of the 12 anti-NY-ESO-1 antibodypositive sera were shown to react. This epitope is located between amino acids 96 and 125, with a core region to be assigned to p106-115 (Fig. 1). According to the predictions of Zeng et al., 1 this region has only a moderate surface accessibility score and a high probability for buried segments. Interestingly, neither of the 2 B-cell epitopes overlaps with known CD8 1 or CD4 1 T-cell epitopes. 2,3 In summary, our results show that important information can be missed when B-cell epitope analysis is confined to antigenic regions predicted by computer-aided algorithms. While computer-aided predictions are good, experimental confirmation is better.