|
|
@@ -23492,29 +23492,56 @@ toolbox
|
|
|
This prepared me to handle the idiosyncracies of any new data set, even
|
|
|
when the new data has problems that I have not previously encountered in
|
|
|
any other data set.
|
|
|
+
|
|
|
\end_layout
|
|
|
|
|
|
-\begin_layout Itemize
|
|
|
-Pipelines are for established processes, not research
|
|
|
-\end_layout
|
|
|
+\begin_layout Standard
|
|
|
+Reusable pipelines have their place, but that place is in automating established
|
|
|
+ processes, not researching new science.
|
|
|
+ For example, the custom
|
|
|
+\begin_inset Flex Glossary Term
|
|
|
+status open
|
|
|
|
|
|
-\begin_layout Itemize
|
|
|
-Research data analysis must be exploratory and flexible.
|
|
|
- Learn the properties of the data and design the analysis to handle them.
|
|
|
+\begin_layout Plain Layout
|
|
|
+fRMA
|
|
|
\end_layout
|
|
|
|
|
|
-\begin_layout Standard
|
|
|
-\begin_inset Flex TODO Note (inline)
|
|
|
+\end_inset
|
|
|
+
|
|
|
+ vectors developed in Chapter
|
|
|
+\begin_inset CommandInset ref
|
|
|
+LatexCommand ref
|
|
|
+reference "chap:Improving-array-based-diagnostic"
|
|
|
+plural "false"
|
|
|
+caps "false"
|
|
|
+noprefix "false"
|
|
|
+
|
|
|
+\end_inset
|
|
|
+
|
|
|
+, are being incorporated into an automated pipeline for diagnosing transplant
|
|
|
+ rejection using biopsy and blood samples from transplant recipients.
|
|
|
+ Once ready, this diagnostic method will consist of normalization using
|
|
|
+ the pre-trained
|
|
|
+\begin_inset Flex Glossary Term
|
|
|
status open
|
|
|
|
|
|
\begin_layout Plain Layout
|
|
|
-This isn't done, but my hands are done for the day.
|
|
|
-
|
|
|
+fRMA
|
|
|
\end_layout
|
|
|
|
|
|
\end_inset
|
|
|
|
|
|
-
|
|
|
+ vectors, followed by classification of the sample by a pre-trained classifier,
|
|
|
+ which outputs a posterior probability of acute rejection.
|
|
|
+ This is a perfect use case for a proper pipeline: repeating the exact same
|
|
|
+ sequence of analysis steps many times.
|
|
|
+ The input to the pipeline is sufficienrtly well-controlled that we can
|
|
|
+ guarantee it will satisfy the assumptions of the pipeline.
|
|
|
+ But research data is not so well-controlled, so when analyzing data in
|
|
|
+ a research context, the analysis must conform to the data, rather than
|
|
|
+ trying to force the data to conform to a preferred analysis strategy.
|
|
|
+ That means having a toolbox of composable methods ready to respond to the
|
|
|
+ observed properties of the data.
|
|
|
\end_layout
|
|
|
|
|
|
\begin_layout Standard
|
