The comparative binding affinities for poly(dA-dT) and poly(dG-dC) of novel antitumor anthracyclines are reported. The data concern, besides the parent compound adriamycin (ADM), 4-demethoxy 6-deoxy 6-aminodaunomycin (II), 9-deoxy-ADM (III), 4-demethyl-6-O-methyl-ADM (IV), and 3'-deamino-3'-hydroxy-4'-epi-ADM (IV). Theoretical computations are performed in parallel for their comparative binding affinities to model double-stranded hexanucleotides, d(GCGCGC)2, d(TATATA)2, and d(CGTACG)2, using the SIBFA (sum of interactions between fragments computed ab inito) procedure. The computations reproduce in a very satisfactory manner the most salient features of the experimental comparative binding affinities. These encompass, in particular, a higher affinity for the d(TATATA)2 oligomer of II than that of ADM, despite the absence of the 14-OH substituent in II, a marked reversal of the CG versus TA sequence selectivity of the neutral compound V, favoring the d(CGCGCG)2 oligomer over the d(TATATA)2 one; and the deleterious effect incurred on the binding affinities by the presence of an O-methyl substituent at position 6 of the chromophore.