Embedding R in Windows applications, and executing R remotely

Deployment of validated statistical algorithms within an end-user’s application
Dr. Mark Demesmaeker, Spotfire Inc., www.spotfire.com
Modern l i f e sc ience research data demands sophis t i c a t e d analys i s . New
techno l og i es and new algor i t hm i c developments requ i r e c lose
col l abo ra t i o n between research - and sta t i s t i c a l organ i za t i o n s . While
once i t was poss ib l e fo r sta t i s t i c i a n s and analy t i c exper t s to work
one- on- one wi th researche r s and the i r data sets , the pace of research
now demands more rap id and broader deployment of analys i s sof twa re to
share analy t i c exper t i s e across the organ i za t i o n . Inc reas i ng l y
sta t i s t i c a l sta f f s are re l y i ng on the i r abi l i t y to in f l u ence the
sof tware used by researche r s as the i r means of ensur i ng r i go rous
val i da t i o n of exper imenta l f i nd i ngs . Unfor t una te l y , of f - the- shel f
sof tware ra re l y conta i ns the sta t i s t i c i a n ’ s algor i t hm or sta t i s t i c a l
method of choice . Stat i s t i c i a n s f i nd themselves becoming appl i c a t i o n
developers in order to prov i de the necessary analy t i c s to the i r enduser
col l eagues . End- users st r ugg l e wi th the complex i t y of analys i s
appl i c a t i o n s , used only occas iona l l y but at cr i t i c a l junc tu r es , to
val i da t e the i r f i nd i ngs . This misa l i g nment of resources can decrease
the ef f ec t i v eness of organ i za t i o n s bet te r prepared to der i ve algor i t hms
than deve lop i ng and mainta i n i ng end- user sof tware .
Spot f i r e Decis i onS i t e i s a v isua l , analy t i c appl i c a t i o n fo r dynamic ,
mult i - dimens iona l data analys i s . Decis i onS i t e is high l y conf i g u r ab l e
and suppor t s guided analys i s in the contex t of any customer process and
data source us ing analy t i c a l too l s and bui l t - in data access
capabi l i t i e s .
Use Decis i onS i t e and the R env i r onment to in t e r ac t i v e l y crea te and tes t R
scr i p t s . These scr i p t s use standard R code and can be parameter i z ed . Using
Decis i onS i t e ’ s suppor t fo r guided analys i s , users deploy R scr i p t s as spec i f i c
end- user data analys i s appl i c a t i o n s . Spot f i r e Decis i onS i t e users run spec i f i c
data analys i s appl i c a t i o n s crea ted and deployed to l i n k with an R server . Users
are prompted fo r and ente r va lues fo r the spec i f i c parameters . Resul t s are
ca lcu l a t ed on the R serve r and re tu r ned to Decis i onS i t e .
The Spot f i r e Advantage Solu t i on fo r R prov i des organ i za t i o ns the
abi l i t y to deploy algo r i t hms deve loped in R wi th i n a Decis i onS i t e
analy t i c appl i c a t i o n .
While many sta t i s t i c a l too l s are f ree l y avai l ab l e , approp r i a t e use of
them may be qui t e di f f i c u l t fo r ind i v i d ua l users . By l i n k i ng
Decis i onS i t e to an R serve r th rough use of the Spot f i r e Advantage
Solu t i on fo r R, the customer ’ s analy t i c sta f f can def i ne spec i f i c
analys i s rou t i nes in R and eas i l y make them avai l a b l e to the res t of
the organ i za t i o n—fo r example , var i ous micro - ar ray normal i z a t i o n
processes avai l a b l e th rough Bioconduc to r . o r g or a novel high th roughput
screen ing
val i da t i o n rou t i n e deve loped by in t e r na l sta t i s t i c i a n s .