The talk will cover two topics: first, to investigate whether cytosine decay resulted from cytosine deamination is a biological selective force shaping codon usage bias in influenza viruses and second, to model haemagglutinin’s (HA) antigenicity using amino acid substitution information. To the first topic, a computational platform has been deployed to compute all the 48 dinucleotide substitution models between two viral RNA sequences. Using the eight RNA sequences from about 2,000 human and avian influenza viruses, we found that CpG deficiency is indeed the most evident bias. In addition, we are the first group to demonstrate that other dinucleotides, such as ApC and ApG also exhibit high avoidance rate. Both observations can be seen on all RNA segments, suggesting that C→T (or G→A on the reversed strain) mutations appear to be subject to a more dominant selection force compared to other mutations. We also show that a phylogenetic reconstruction based on the tendency of CpG, ApC or ApG avoidance produces trees that are consistent with the one built from traditional maximum likelihood estimation using amino acid sequence information, another observation indicating that cytosine deamination could play an important role in virus evolution. For the second topic, using a training and a testing sets containing 249 and 208, respectively, pairs of HA inhibition data and a multi-resolution analysis, we have built a computational system that predicts whether an unknown amino acid substitution might result in an evident antigenic drift. The sensitivity and specificity are 0.85 and 0.87, respectively, for the independent testing data. Among the 20 identified immuno-dominant sites along the HA sequence, eight sites have been reported in literatures and seven are located on known epitope regions. We also found that epitope B exhibit a slightly stronger prediction power, suggesting that substitutions on this epitope might be more relevant to antigenic drift, also reported in a recently published literature.