Statistical Genetics, Kyoto University

• Paper link ペイパーリンク
  • PEER(Using probabilistic estimation of expression residuals (PEER) to obtain increased power and interpretability of gene expression analyses)(Nat Protoc. 2012 Feb 16; 7(3): 500–507)
  • VBQTL(A Bayesian Framework to Account for Complex Non-Genetic Factors in Gene Expression Levels Greatly Increases Power in eQTL Studies)(PLoS Comput Biol. 2010 May 6;6(5):e1000770. doi: 10.1371/journal.pcbi.1000770.)
• 概要 Abstract
  • PEERはVBQTLを用いて、体系的発現解析データに潜む隠れた因子(Surrogate factors)を取り出したうえで、それを取り除いてeQTL mappingをしたり、その因子を使いながらeQTLマッピングをしたりする、アプリケーション
  • PEER is a software package based on VBQTL for eQTL: VBQTL identifies hidden (surrogate factors) from expression profiles and PEER uses the surrogate factors to perform eQTL mapping with or without incorporating the hidden factors.
  • そのダウンロードはR package,Python interface , where it is downloadable from.
  • サロゲート因子には細胞タイプとその割合、実験条件、細胞の遺伝的バリアント条件などが含まれる
  • Cell types and their fraction, experimental conditions, genetic variants and others are among the surrogate factors.
  • VBQTL は観察されていないけれど、発現プロファイルに影響を与えている因子(サロゲート因子)を見出す方法の一つ
  • VBQTL is one way to identify surrogate factors, which are not observed but are affecting expression profiles.
  • VQTLがどういうものかを理解するには、PLoS Comput BiolのMethodsに書かれている以下の部分の諸単語がわかればよいだろう。リンク先の説明文書は長いけれど、冒頭(等)の短いパラグラフで概要をつかむだけで十分。
  • To understand VQTL, it is almost adequate to understand words in the following paragraph in Methods section of PLoS Comput Biol. The linked documents are long but read header paragraph of each would be enough.

We perform Bayesian inference(En) ベイズ推定(Jp) in the joint model(Joint probability (En),同時分布 (Jp)), which is appealing for
several reasons. First, it allows possible dependencies between the
different sources of variation to be captured. The effects of the
genotype, known and hidden factors are learned jointly, taking other
parts of the model into account. Propagation of uncertainty(En) 確率伝搬法(Jp) leads to
more accurate parameter estimates, and avoids possible pathologies (Pathological in Math/Computer Sciences(En) Pathological in Math/Computer Sciences(Jp),
for instance of maximum likelihood methods (Pathological with MLE(En)). Second, Bayesian
inference allows different models to be flexibly combined according to
the needs of a particular study. Many existing approaches can be cast as
special cases of this general framework, with some examples given in
Figure 1 (Look at figure 1 and find three analysis frameworks). Finally, the Bayesian approach leads itself to efficient
approximate inference schemes such as variational methods (Bayesian network is one of Graphical models(En))Bayesian network is one of Graphical models(Jp)) Variational models(En) ,
rendering the resulting algorithms applicable to large-scale and high-
dimensional datasets. Also, variational learning allows an inference
schedule to be specified by the user, leading to distinct algorithms
with different computational complexity and properties.