N6-methyladenine (6mA) is a rediscovered DNA modification in eukaryotic genomes. To explore the distribution and functions of 6mA, it is of paramount option to use immunoprecipitation to select 6mA-containing DNA fragments for genome-wide sequencing. Presumably, most of the 6mA-free fragments are removed, and the copulling down of the residual is stochastic and sequence-independent and thus they should not be called as peaks by computation. Surprisingly, here we show the predominance of 6mA-free fragments in the pulled-down fractions. By taking advantage of the submicromolar affinity of the antibodies, we further develop an elegant, multiple-round immunoprecipitation (MrIP) approach and show that 6mA-containing fragments can be enriched over 9100-fold and dominate in the final pulled-down fractions. This biochemical approach would greatly reduce the peak calling bias, which is caused by handling of dominated 6mA-free DNA fragments with an assumption-based algorithm computation and facilitates 6mA-pertinent data mining. The MrIP concept is extendable for the genome-wide sequencing of diverse DNA modifications.