11.07.2015 Views

2DkcTXceO

2DkcTXceO

2DkcTXceO

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

192 Journey into genetics and genomicsand genomics. Looking back, a good timing and a stimulating collaborativeenvironment made my transition easier. In the mean time, moving into a newfield with limited background requires patience, courage, and willingness tosacrifice, e.g., having a lower productivity in the first few years, and moreimportantly, identifying a niche.18.3 A few lessons learnedImportance to be a scientist besides a statistician: While working onstatistical genetics in the last few years, an important message I appreciatemore and more is to be a scientist first and then a statistician. To make aquantitative impact in the genetic field, one needs to be sincerely interestedin science, devote serious time to learn genetics well enough to identify importantproblems, and closely collaborate with subject-matter scientists. It isless a good practice to develop methods first and then look for applications ingenetics to illustrate the methods. By doing so, it would be more challengingto make such methods have an impact in real world practice, and it is morelikely to follow the crowd and work on a problem at a later and more maturedstage of the area. Furthermore, attractive statistical methods that arelikely to be popular and advance scientific discovery need to integrate geneticknowledge well in method development. This will require a very good knowledgeof genetics, and identifying cutting-edge scientific problems that requirenew method development, and developing a good sense of important and lessimportant problems.Furthermore, the genetic field is more technology-driven than many otherhealth science areas, and technology moves very fast. Statistical methods thatwere developed for data generated by an older technology might not be applicablefor data generated by new technology. For example, normalizationmethods that work well for array-based technology might not work well forsequencing-based technology. Statistical geneticists hence need to closely followtechnological advance.Simple and computationally efficient methods carry an importantrole: To make analysis of massive genetic data feasible, computationallyefficient and simple enough methods that can be easily explained to practitionersare often more advantageous and desirable. An interesting phenomenon isthat simple classical methods seem to work well in practice. For example, inGWAS, simple single SNP analysis has been commonly used in both the discoveryphase and the validation phase, and has led to discovery of hundreds ofSNPs that are associated with disease phenotypes. This presents a significantchallenge to statisticians who are interested in developing more advanced andsophisticated methods that can be adopted for routine use and outperform

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!