OReilly Media. - Bioinformatics data skills: [reproducible and robust research with open source tools]

Many biologists starting out in bioinformatics tend to equate learningbioinformatics with learning how to run bioinformatics software. This is anunfortunate and misinformed idea of what bioinformaticians actually do. This isanalogous to thinking learning molecular biology is just learningpipetting. Other than a few simple examples used to generate data in, this book doesnt cover running bioinformatics software like aligners, assemblers, or variant callers. Running bioinformatics software isnt all that difficult, doesnt take much skill, and it doesnt embody any of the significant challenges of bioinformatics. I dont teach how to run these types of bioinformatics applications in Bioinformatics Data Skills for the following reasons:

• Its easy enough to figure out on your own

• The material would go rapidly out of date as new versions of software or entirely new programs are used in bioinformatics

• The original manuals for this software will always be the best, most up-to-date resource on how to run a program

Instead, the approach of this book is to focus on the skills bioinformaticiansuse to explore and extract meaning from complex, large bioinformatics datasets.Exploring and extracting information from these datasets is the fun part ofbioinformatics research. The goal of Bioinformatics Data Skills is to teachyou the computational tools and data skills you need to explore these largedatasets as you please. These data skills give you freedom; youll be able tolook at any bioinformatics datain any format, and files of any sizeand begin exploring data to extract biological meaning.

Throughout Bioinformatics Data Skills, I emphasize working in a robust andreproducible manner. I believe these two qualitiesreproducibility androbustnessare too often overlooked in modern computational work. Byrobust, I mean that your work is resilient against silent errors, confounders,software bugs, and messy or noisy data. In contrast, a fragile approach is onethat does not decrease the odds of some type of error adversely affecting yourresults. By reproducible, I mean that your work can be repeated by other researchersand they can arrive at the same results. For this to be the case, your work must be well documented, and your methods, code, and data all need to be available so that other researchers have the materials to reproduce everything. Reproducibility also relieson your work being robustif a workflow run on a different machine yields adifferent outcome, it is neither robust nor fully reproducible. I introduce theseconcepts in more depth in , and these are themes that reappearthroughout the book.