Hello peers,
I'm looking for papers/tutorials/information on how to classify transcripts based on species. For example, if I was going to be sequencing a small anemone doing a whole-body RNA extraction I want to be able to differentiate between the anemone and the food it has eaten (bacteria, plankton, etc.).
I know that the Lehnert et al. 2014 (Extensive Differences in Gene Expression between Symbiotic and Aposymbiotic Cnidarians) designed a custom Support Vector Machine algorithm "TopSort"; however, it is not open-source.
If anyone could provide other examples (preferably with open-source software that would require no, or very little, tweaking) It would be most appreciated. I think that gut microbiome or metagenomic research would cover this but I haven't found anything yet.
Regards,
Matthew J. Oldach
LIBSVM and LIBLINEAR are two popular open source machine learning libraries. Written in C++ though with a C API.
Ref: http://en.wikipedia.org/wiki/LIBSVM
http://www.csie.ntu.edu.tw/~cjlin/libsvm/
http://www.csie.ntu.edu.tw/~cjlin/liblinear/
LIBSVM is an SVM library and will not solve your problem of course.
If don't find something I am willing to collaborate and come up with a solution for your data that we can publish as well.
Kraken is the closest thing I've come across, but doesn't use any fancy machine learning I think. It looks at each kmer in the sequence in question and asks whether that particular kmer has been seen before, only in a particular clade, and thus, can the sequence be classified to a particular clade. In this way it reduces your set of input sequences to a set of diagnostic features/markers that can be used to classify each sequence to a clade.
You could build a machine learning approach around this I guess, by training on known genomes, using each kmer as a feature, and deriving a classifier model relating the predictive value of each kmer in relation to each taxon.
Maybe SVM is only a nicety here though, and you could make use of the existing (open) kraken software. You would probably need to make your own kraken database, enriched for Cnidarian sequences (the default kraken is bacterial as far as I know), and bear in mind that this is a job for a machine with very large RAM.
Bonus points for returning the kraken to the underwater world.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Bioinformatics
- Bioinformatic Tools