Modern Statistics For Modern Biology

Cambridge University Press978-1-108-70529-5 — Modern Statistics for Modern BiologySusan Holmes , Wolfgang HuberFrontmatterMore InformationModern Statistics for Modern BiologyIf you are a biologist and want to get the best out of the powerful methodsof modern computational statistics, this is your book. You can visualize andanalyze your own data, apply unsupervised and supervised learning, integrate datasets, apply hypothesis testing, and make publication-quality figures using the power of R/Bioconductor and ggplot2.This book will teach you ‘‘cooking from scratch’,’ from raw data to beautifulilluminating output, as you learn to write your own scripts in the R languageand to use advanced statistics packages from CRAN and Bioconductor. It covers a broad range of basic and advanced topics important in the analysis ofhigh-throughput biological data, including principal component analysis andmultidimensional scaling, clustering, multiple testing, unsupervised and supervised learning, resampling, the pitfalls of experimental design, and powersimulations using Monte Carlo, and it even reaches networks, trees, spatialstatistics, image data, and microbial ecology. Using a minimum of mathematical notation, it builds understanding from well-chosen examples, simulation, visualization, and above all hands-on interaction with data and code. R package msmb contains complete code and the example datasets, allowing students to recreate all examples, figures, and results in the book Solutions, slides, and dynamic material available on the course website Introduces methods on a ‘‘need to know’’ basis, so students tackle biological questions immediately and understand motivation for the methods Real-life examples done from scratch, guiding students through realisticcomplexities and building practical intuition Includes a wrap-up chapter that explains the complete workflow from design of experiments to analysis of results, identifying common pitfallswith big data All figures and results generated by the code in the book, demonstratinghow reproducible research worksSUSAN HOLMES is Professor of Statistics at Stanford University, California.She specializes in exploring and visualizing multidomain biological data, using computational statistics to draw inferences in microbiology, immunologyand cancer biology. She has published over 100 research papers, and has beena key developer of software for the multivariate analyses of complex heterogeneous data. She was the Breiman Lecturer at NIPS 2016, has been named aFields Institute fellow, and is currently a fellow at the Center for the AdvancesStudy of the Behavioral Sciences.WOLFGANG HUBER is Research Group Leader and Senior Scientist at theEuropean Molecular Biological Laboratory, where he develops computationalmethods for new biotechnologies and applies them to biological discovery.He has published over 150 research papers in functional genomics, cancerand statistical methods. He is a founding member of the open-sourcebioinformatics software collaboration Bioconductor and has co-authored twobooks on Bioconductor. in this web service Cambridge University Presswww.cambridge.org

Cambridge University Press978-1-108-70529-5 — Modern Statistics for Modern BiologySusan Holmes , Wolfgang HuberFrontmatterMore Information in this web service Cambridge University Presswww.cambridge.org

Cambridge University Press978-1-108-70529-5 — Modern Statistics for Modern BiologySusan Holmes , Wolfgang HuberFrontmatterMore InformationModernStatistics forModernBiologySusan HolmesStanford University, CaliforniaWolfgang HuberEuropean Molecular Biology Laboratory in this web service Cambridge University Presswww.cambridge.org

Cambridge University Press978-1-108-70529-5 — Modern Statistics for Modern BiologySusan Holmes , Wolfgang HuberFrontmatterMore InformationUniversity Printing House, Cambridge CB2 8BS, United KingdomOne Liberty Plaza, 20th Floor, New York, NY 10006, USA477 Williamstown Road, Port Melbourne, VIC 3207, Australia314–321, 3rd Floor, Plot 3, Splendor Forum, Jasola District Centre, New Delhi – 110025, India79 Anson Road, #06–04/06, Singapore 079906Cambridge University Press is part of the University of Cambridge.It furthers the University’s mission by disseminating knowledge in the pursuit ofeducation, learning, and research at the highest international levels of excellence.www.cambridge.orgInformation on this title: www.cambridge.org/9781108705295DOI: 10.1017/9781108551441 Susan Holmes and Wolfgang Huber 2018This publication is in copyright. Subject to statutory exceptionand to the provisions of relevant collective licensing agreements,no reproduction of any part may take place without the writtenpermission of Cambridge University Press.First published 2018Printed and bound in Great Britain by Clays Ltd, Elcograf S.p.A.A catalogue record for this publication is available from the British Library.Library of Congress Cataloging-in-Publication DataISBN 978-1-108-70529-5 PaperbackAdditional resources for this publication at www.cambridge.org/msmbCambridge University Press has no responsibility for the persistence or accuracyof URLs for external or third-party internet websites referred to in this publicationand does not guarantee that any content on such websites is, or will remain,accurate or appropriate.Image credits for chapter openers: Chapter 1, Wikicommons;Chapter 4, xkcd.com/1347; Chapter 5, mikedabell/iStock/Getty Images;Chapter 6, extract from xkcd.com/882/; Chapter 7, The Matrix: scene 291 Close onComputer Screen Warner Bros.; Chapter 8, xkcd.com/1725;Chapter 9, Robert Orchard/Moment/Getty Images;Chapter 13, University of Adelaide Library: Rare Books and SpecialCollections, R.A. Fisher Digital Archive,http://hdl.handle.net/2440/81670. in this web service Cambridge University Presswww.cambridge.org

Cambridge University Press978-1-108-70529-5 — Modern Statistics for Modern BiologySusan Holmes , Wolfgang HuberFrontmatterMore InformationFor Sonia, Sara, Agnès, Johnny, Camille. . . and the “girls” who make me love the life sciencesFor Alexander in this web service Cambridge University Presswww.cambridge.org

Cambridge University Press978-1-108-70529-5 — Modern Statistics for Modern BiologySusan Holmes , Wolfgang HuberFrontmatterMore Information in this web service Cambridge University Presswww.cambridge.org

Cambridge University Press978-1-108-70529-5 — Modern Statistics for Modern BiologySusan Holmes , Wolfgang HuberFrontmatterMore InformationContentsIntroductionxvii1Generative Models for Discrete Data12Statistical Modeling193High-Quality Graphics in R534Mixture Models835Clustering1076Testing1397Multivariate Analysis1618High-Throughput Count Data1919Multivariate Methods for Heterogeneous Data21710 Networks and Trees24911 Image Data27912 Supervised Learning30913 Design of High-Throughput Experiments and Their Analyses337AcknowledgementsBibliographyStatistical ConcordanceIndex365367377379 in this web service Cambridge University Presswww.cambridge.org

Cambridge University Press978-1-108-70529-5 — Modern Statistics for Modern BiologySusan Holmes , Wolfgang HuberFrontmatterMore InformationExpanded ContentsIntroductionxviiWhat is happening in biological data analysis?1The challenge: heterogeneityxviiWhat’s in this book?xviiiComputational tools for modern biologistsxxWhy R and Bioconductor?xxiHow to read this bookxxiiGenerative Models for Discrete Data11.1Goals for this chapter11.2A real example11.3Using discrete probability models21.3.1Bernoulli trials31.3.2Binomial success counts41.3.3Poisson distributions51.3.4A generative model for epitope detection1.426Multinomial distributions: the case of DNA101.4.111Simulating for power1.5Summary of this chapter151.6Further reading151.7Exercises16Statistical Modeling192.1Goals for this chapter192.2The difference between statistical and probabilistic models202.3A simple example of statistical modeling202.3.1242.42.5Classical statistics for classical dataBinomial distributions and maximum likelihood252.4.125An exampleMore boxes: multinomial data272.5.1DNA count modeling: base pairs27Nucleotide bias272.5.22.6 in this web service Cambridge University Pressxvii2The χ distribution292.6.129Intermezzo: quantiles and the quantile–quantile plotwww.cambridge.org

Cambridge University Press978-1-108-70529-5 — Modern Statistics for Modern BiologySusan Holmes , Wolfgang HuberFrontmatterMore Informationexpanded contents ix2.72.82.92.102.112.122.133Chargaff’s Rule2.7.1Two categorical variables2.7.2A special multinomial: Hardy–Weinberg equilibrium2.7.3Concatenating several multinomials: sequence motifs and logosModeling sequential dependencies: Markov chainsBayesian thinking2.9.1Example: haplotype frequencies2.9.2Simulation study of the Bayesian paradigm for the binomialExample: occurrence of a nucleotide pattern in a genome2.10.1 Modeling in the case of dependenciesSummary of this chapterFurther readingExercisesHigh-Quality Graphics in R3.1 Goals for this chapter3.2 Base R plotting3.3 An example dataset3.4 ggplot23.4.1Data flow3.4.2Saving figures3.5 The grammar of graphics3.6 Visualizing data in 1D3.6.1Barplots3.6.2Boxplots3.6.3Violin plots3.6.4Dot plots and beeswarm plots3.6.5Density plots3.6.6ECDF plots3.6.7The effect of transformations on densities3.7 Visualizing data in 2D: scatterplots3.7.1Plot shapes3.8 Visualizing more than two dimensions3.8.1Faceting3.8.2Interactive graphics3.9 Color3.10 Heatmaps3.10.1 Dendrogram ordering3.10.2 Color spaces3.11 Data transformations3.12 Mathematical symbols and other fonts3.13 Genomic data3.14 Summary of this chapter3.15 Further reading3.16 Exercises in this web service Cambridge University bridge.org

Cambridge University Press978-1-108-70529-5 — Modern Statistics for Modern BiologySusan Holmes , Wolfgang HuberFrontmatterMore Informationxexpanded contents4Mixture Models834.1834.25Goals for this chapterFinite mixtures844.2.1Simple examples and computer experiments844.2.2Discovering the hidden group labels864.2.3Models for zero-inflated data904.2.4More than two components914.3Empirical distributions and the nonparametric bootstrap924.4Infinite mixtures944.4.1Infinite mixture of normals944.4.2Infinite mixtures of Poisson variables964.4.3Gamma distribution: two parameters (shape and scale)964.4.4Variance-stabilizing transformations994.5Summary of this chapter1024.6Further reading1034.7Exercises104Clustering1075.1Goals for this chapter1075.2What are the data and why do we cluster them?1085.2.11085.35.45.55.6How do we measure similarity?1105.3.11125.85.9 in this web service Cambridge University PressComputations related to distances in RNonparametric mixture detection1135.4.1k -methods: k -means, k -medoids and PAM1135.4.2Tight clusters with resampling114Clustering examples: flow cytometry and mass cytometry1155.5.1Flow cytometry and mass cytometry1155.5.2Data preprocessing1165.5.3Density-based clustering118Hierarchical clustering5.6.15.7Clustering can sometimes lead to discoveries120How to compute (dis)similarities between aggregated clusters? 120Validating and choosing the number of clusters1235.7.1Using the gap statistic1255.7.2Cluster validation using the bootstrap127Clustering as a means for denoising1295.8.1Noisy observations with different baseline frequencies1305.8.2Denoising 16S rRNA sequences1315.8.3Inferring sequence variants132Summary of this chapter1345.10 Further reading1345.11 Exercises135www.cambridge.org