To elucidate the adaptive strategies of influenza A viruses (IAVs) to human, we proposed a computational approach to identify human-adaptive mutations in seasonal IAVs, which have not been analyzed comprehensively. We compared representative PB2 sequences of 1425 avian IAVs and 2176 human IAVs and identified a total of 42 human-adaptive markers, including 28 and 31 markers in PB2 proteins of seasonal viruses H1N1 and H3N2, respectively. Notably, this comprehensive list encompasses almost all the markers identified in prior computational studies and 21 novel markers including an experimentally verified mutation K526R, suggesting the predictive power of our method. The strength of our analysis derives from the enormous amount of recently available sequences as well as the recognition that human-adaptive mutations are not necessarily conserved across subtypes. We also utilized mutual information to profile the inter-residue coevolution in PB2 protein. A total of 35 and 46 coevolving site pairs are identified in H1N1 and H3N2, respectively. Interestingly, 13 out of the 28 (46.4%) identified markers in H1N1 and 16 out of the 31 (51.6%) in H3N2 are embraced in the coevolving pairs. Many of them are paired with well-characterized human-adaptive mutations, indicating potential epistatic effect of these coevolving residues in human adaptation. Additionally, we reconstructed the PB2 evolutionary history of seasonal IAVs and demonstrated the distinct adaptive pathway of PB2 segment after reassortment from H1 to H3 lineage. Our study may provide clues for further experimental validation of human-adaptive mutations and shed light on the human adaptation process of seasonal IAVs.