A procedure involving several elements from biology, artificial intelligence, statistics, and mathematics to forecast influenza A variants to a very short term is proposed. This new mechanism tackles the huge complexity of this kind of forecast in a systematic way. Results show forecasts of both recurrent and new mutations. This procedure would also allow the identification of potential mutations of certain hemagglutinin and neuraminidase proteins, since its core is based on mining a number of protein relationships involving hemagglutinin and neuraminidase mutations. The findings of these mutations of the influenza A viruses that will circulate each season are essential to develop effective vaccines, since the rapid evolution of viruses is their main way of evading human immunity. The results show that the proposed procedure is promising to further improve this type of analysis strategy and, therefore, to discover true relationships giving true mutations.
