Probe to Probe Mapping
Neil Saunders highlighted a problem in software that is supposed to work with different array types, CNAmet:"the problem of mapping measurements between different array types is not dealt with by CNAmet": or anything else in my experience (link)There is a software package that does these type of interconversions, Absolute ID Convert, which uses genomic coordinate mapping to interconvert between array IDs. The paper was published in BMC Bioinformatics last year (link) by a PhD student from my group, who has gone on to PostDoc at Harvard.
Sequencing Errors due to Sample Processing
Nick Loman (twitter, blog) has an interesting post highlighting recent developments in errors in Next-Generation Sequencing arising from sample preparation protocols. Casey Bergman is maintaining a collection of papers related to sequencing errors.
Protein Annotation Biases
Iddo Friedberg has a new arXiv preprint examining the effect of biases in the experimental annotations of protein function, and their effect on our understanding of protein function. I haven't read it yet, but from the abstract, it looks like it is definitely something to check out.
Free Statistical Methods Ebook
If you work with any amount of data (likely if you read this blog regularly), then you might be interested in the free ebook of "The Elements of Statistical Learning" (Hastie, Tibshirani, and Friedman), which is a book on a wide variety of machine learning methods, written by experts in the field. I downloaded my copy, you should too.
I should probably expand on my tweet in a blog post. Mapping probe IDs between array platforms is often "relatively easy." I'm more concerned with how to summarize different measurements on a "per gene" basis. As in all those R packages and associated papers on data integration where "rows are genes and columns are samples", with no explanation as to how this is done.
ReplyDeleteSorry for my misunderstanding Neil. That is a big problem, that lots of papers and tutorials seem to just assume it is self-evident. Would be good to have more people be explicit on this.
ReplyDeleteHowever, regarding mapping between platforms, I would argue that this is something that isn't that easy, and that lots of bad assumptions have been made in how this is done to date, which is why our group developed AbsIDConvert.