Prediction of antigenic evolution of influenza viruses through Bayesian estimation using statistics on strings

2019 Fiscal Year Final Research Report

• Project/Area Number: 16H02863
• Research Category: Grant-in-Aid for Scientific Research (B)
• Allocation Type: Single-year Grants
• Section: 一般
• Research Field: Intelligent informatics
• Research Institution: Hokkaido University
• Principal Investigator: Ito Kimihito 北海道大学, 人獣共通感染症リサーチセンター, 教授 (60396314)
• Project Period (FY): 2016-04-01 – 2019-03-31
• Keywords: インフルエンザウイルス / 抗原変異 / データ同化 / 粒子フィルタ

Outline of Final Research Achievements

By integrating the quasi spices theory with infectious disease modelling, we developed a mathematical model of viral sequences, infection, and herd immunity. Actual observations of viral sequences and the number of reported cases of H1N1 2009 pandemic viruses were integrated into computer simulations, and parameters in the model were estimated by the data assimilation technique called the particle filter. For every 30 days 100,000 simulations were ran and filter distributions of parameters were calculated using amino acid sequences observed in the corresponding periods. Six-month period predictions were made using the estimated parameters. As a result, the developed method predicts the actual amino acid substitutions on the hemagglutinin molecule with a precision ratio of 79% and recall ratio of 53% .

Free Research Field

計算機科学

Academic Significance and Societal Importance of the Research Achievements

インフルエンザの予防にはワクチン接種が有効であるが，人の免疫圧による選択淘汰を受けてウイルスの遺伝子が変異し続けるため， ワクチン株を頻繁に更新しなければならない。そこで，本研究では，ワクチン株を先回りして準備するために，感染症数理疫学と集団遺伝学を融合し，ウイルスの遺伝子配列の文字列統計量から，感染症流行モデルのパラメータを推定する手法を開発し，その予測精度を明らかにした。本手法は 6カ月後のアミノ酸置換を，適合率79%，再現率53%で予測できることが明らかになり，今後のワクチン政策への応用が期待される。