A Nice talk on software innovations for CMB
Robert Gentleman - Harvard Biostatistics-
One of the pioneers of R and Bioconductor.
Software Innovation for computational biology and bioinformatics
- range of problems
methods for sequences,microarray,interactions,modelling,machine learning,
- Famous architect.
- More famous yesterday and today and more tomorrow and the day next.
- I am completely confounded by the idea of software reuse.
- To promote reproducible research.
ot hard.
1) Reduce complexity. Make life easy
2) Should provide basis of simulation.
3) Language sfor describing models.
AI NOT EQUAL TO i
COMPONENET SOFTWARE
1) SHOULD HAVE A CLEAR FUNCTION
2) A well defined set of inputs and ops
3) should be versioned
4) should have unit testing
5) should be automatically available
6) should be self documenting
- R Packages satisfy most of thest requirements.
- It is quite clear well defined web services can play a similar role. Given a well defined API one can use it as a module.
- FEW WEB SERVICES AVAIALBLE
- if you make a publication about computation The reader has to believe that you have done a whole bunch of research right.
- Compendium : Give users a compendium.
www.bioconductor.org/Docs/Papers/2003/Compendium.
- Use makefiles to
- plots should be reproducible!!!!!
- Literate Programming. Don Knuth.
- Rearrange computer programs to have people make sense.
- Literate Data Analysis.
- Sweave. by Fritz Rodges.
- Compendium: text, code, data
- I have document and I have transformer, I can make variety of different outputs from same input
- Might want to checkout Bioconductor for Protein Protien interactions.
- RBGL package
- Boost Library
- Rendering Layout
rgraphviz package, www.graphviz.org
- Tandem affinity purificaiton. Taking a protein and like a bait, doing fishing waiting for something to stick to it.
Robert Gentleman - Harvard Biostatistics-
One of the pioneers of R and Bioconductor.
Software Innovation for computational biology and bioinformatics
- range of problems
methods for sequences,microarray,interactions,modelling,machine learning,
- Famous architect.
- More famous yesterday and today and more tomorrow and the day next.
- I am completely confounded by the idea of software reuse.
- To promote reproducible research.
ot hard.
1) Reduce complexity. Make life easy
2) Should provide basis of simulation.
3) Language sfor describing models.
AI NOT EQUAL TO i
COMPONENET SOFTWARE
1) SHOULD HAVE A CLEAR FUNCTION
2) A well defined set of inputs and ops
3) should be versioned
4) should have unit testing
5) should be automatically available
6) should be self documenting
- R Packages satisfy most of thest requirements.
- It is quite clear well defined web services can play a similar role. Given a well defined API one can use it as a module.
- FEW WEB SERVICES AVAIALBLE
- if you make a publication about computation The reader has to believe that you have done a whole bunch of research right.
- Compendium : Give users a compendium.
www.bioconductor.org/Docs/Papers/2003/Compendium.
- Use makefiles to
- plots should be reproducible!!!!!
- Literate Programming. Don Knuth.
- Rearrange computer programs to have people make sense.
- Literate Data Analysis.
- Sweave. by Fritz Rodges.
- Compendium: text, code, data
- I have document and I have transformer, I can make variety of different outputs from same input
- Might want to checkout Bioconductor for Protein Protien interactions.
- RBGL package
- Boost Library
- Rendering Layout
rgraphviz package, www.graphviz.org
- Tandem affinity purificaiton. Taking a protein and like a bait, doing fishing waiting for something to stick to it.
Comments