What's left to do in bioinformatics
Recently, I stumbled across this paper that describes an approach to predict the localization of proteins in yeast.
I was suprised and giggling in my chair, because the whole yeast proteome was screened for its localization by Huh et al. , so why would any body agree to publish such a method. Obviously, the study by Huh et al. fails to get a signal for each and every protein due to experimental reasons. Some proteins are simply not accessible and the authors of the computational study simply fill the gaps.
However, it appears to me that the task feels trivial and wonder what bioinformatics, at least the part that's want to predict biolological properties. So far, no method was really good at predicting the localization from sequence but the task appears simple compared to fold prediction - which has been predicted (by David Baker, if I am not mistaken) to be good once all the folds have been discovered experimentally.
I wonder, whether there will be bioinformatic predictions of real impact. Can you name a few?
I was suprised and giggling in my chair, because the whole yeast proteome was screened for its localization by Huh et al. , so why would any body agree to publish such a method. Obviously, the study by Huh et al. fails to get a signal for each and every protein due to experimental reasons. Some proteins are simply not accessible and the authors of the computational study simply fill the gaps.
However, it appears to me that the task feels trivial and wonder what bioinformatics, at least the part that's want to predict biolological properties. So far, no method was really good at predicting the localization from sequence but the task appears simple compared to fold prediction - which has been predicted (by David Baker, if I am not mistaken) to be good once all the folds have been discovered experimentally.
I wonder, whether there will be bioinformatic predictions of real impact. Can you name a few?
spitshine - 2005-04-08 18:23