Proposal of interactive large scale Bayesian network estimation method and its application to biological data

2018 Fiscal Year Final Research Report

• Project/Area Number: 15K00402
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Life / Health / Medical informatics
• Research Institution: Kansai University (2017-2018); Osaka University (2015-2016)
• Principal Investigator: takenaka yoichi 関西大学, 総合情報学部, 教授 (00324830)
• Project Period (FY): 2015-04-01 – 2019-03-31
• Keywords: 生物情報学 / ベイジアンネットワーク / 遺伝子制御ネットワーク / 一塩基多型 / 大規模データ / 因果関係

Outline of Final Research Achievements

In data analysis, causality that indicates cause and effect between events is more important than correlation that simply indicates that changes between events are similar. Bayesian networks have greatly contributed to small-scale data analysis as a model expressing the causality and a method of estimating the causality. The Bayesian network is a probabilistic model representing causality with conditional probability, and is a non-cyclic directed graph representing causality with directed edges between vertices corresponding to events.

In this study, we proposed a method to expand the number of events that Bayesian networks can estimate and to compare data observed from different sources for the same event. And the effectiveness was clarified through verification using real data in the field of biology.

Free Research Field

生物情報学

Academic Significance and Societal Importance of the Research Achievements

ベイジアンネットワークによる因果関係解析は、生物学、医学、薬学に止まらず医師の診断、イメージ認識、言語認識でも用いられる汎用的な手法である。各分野において今後もデータ産生量が増大し続ける事は確実であろう。そのため、因果関係を推定可能の事象数を拡大可能な手法、および観測データ比較を可能とする手法の需要は今後も拡大し続けるであろう。そのため、本研究の成果は生物学に止まらず、他分野データの解析においても新しい知見を獲得するための手段として活躍する事が期待される。