Statistics - Bioinformatics Computing - page 187

Herman Hollerith's tabulating machine was used to analyze the census—and they changed within
about a year. Buoyed by this success, the Tabulating Machine Company evolved to service other
areas, and eventually became known as IBM.
The U.S. Census Bureau collected samples from the U.S. population in the 1940 census, a modern
statistical technique, instead of collecting data from every citizen. Because even this task was
daunting using manual methods, the census bureau commissioned the UNIVAC Division of Remington
Rand to build the first commercial non-military digital computer in the U.S., a UNIVAC model, in the
late 1940s. Today, many molecular biologists routinely work with data sets that are considerably
larger than those produced by the latest U.S. census, and they use a variety of advanced statistical
techniques to do so.
This chapter explores the practical considerations involved in applying statistical techniques to
modern bioinformatics challenges. It illustrates the range and complexity of issues that arise in
controlling for the variability (which, in this discussion, encompasses errors) associated with
microarray experiments and other bioinformatics work. "Statistical Concepts" introduces the
underlying concepts of randomness and variability, while the " Microarrays " section provides an
overview of the microarray experimental process. The " Imperfect Data " section reveals the numerous
potential sources of variability in microarray experiments.
The " Basics " section relates microarray experiments to fundamental statistical concepts, while
"Quantifying Randomness" discusses how randomness and variability are assigned to devices and
processes. "Data Analysis" discusses how experimental output data are evaluated, and "Tool
Selection" examines the criteria for statistical analysis tool selection. The " Statistics of Alignment "
and " Clustering and Classification " sections illustrate the practical application of statistical concepts.
" On the Horizon " introduces the technological innovations that bioinformatics is pushing forward,
often ahead of the theoretical statistical underpinnings. "Endnote" addresses the implication of
succumbing to the pressure to treat statistics as a black box solution to modern research challenges.
This chapter, like any topic on statistical methods, should be considered a roadmap to potential
issues to consider in discussing the selection of statistical methods with an expert statistician familiar
with bioinformatics issues.