Hands-on Statistical Methods: A Case Study with Hidden Markov Models Using Simulations and Experiments

Abstract

Biological processes are often very complicated compared with physics and chemistry. One of the newest and most challenging interactions between biology and computational science comes from modern molecular biology and bioinformatics, where Hidden Markov Models (HMM) are widely applied tools. This paper presents the background, theory and HMM algorithms based on examples from Gregor Mendel’s classical plant experiments. This approach aims to achieve some intuitive advantages in a biological and bioinformatical setting, because the pedagogy goes from the known to the unknown. It only presumes basic knowledge of genetics, statistics and matrix algebra. The student may gain insight into the complex HMM methodology by running “experiments” with the application MendelHMM in a kind of “digital laboratory”. The optimal model can only be sought in a certain probabilistic sense. This process is known as machine learning.