Thursday, November 03, 2005

Recent reads - two useful applications of bioinformatics

Is bioinformatics actually producing any useful tools or discovering anything new ? I would like to think so :). Here is a table from The Scientist showing the top ten cited papers of the last 2 years, last 10 years and of all time. Blast and Clustal are among the two ten cited papers in the last 10 years and MFold is within the top ten cited papers of the last two years.

Keeping in the spirit of practical applications of computational biology here two recent papers I read.

One is about the computational design of ribozymes. The authors computationally designed different ribozymes that could perform different logical functions. For example they were able to design a AND rybozyme that would self cleave only in the presence of two different particular oligos. They experimentally validated the results in vitro. These ribozymes can be combined to make more complicated circuits and could ultimately be used inside the cells to interfere with the networks in a rational matter or maybe to act as sensors,etc. They don't discuss how applicable this results are for in-vivo studies since ion content, pH and a lot of other things cannot be controlled in the same way.

Another interesting paper is about predicting the specificity of protein-DNA binding using structural models. They did this by developing a model for the free energy of protein-DNA interactions. With the model developed they could calculate the binding energy for structures of proteins bound to DNA and to any such complex after changing the bases in the DNA sites in contact with the protein. This results in a position specific scoring matrix that informs us of what are the preferred nucleotides at each positions for a particular DNA binding protein domain.
The protein-DNA interaction module is incorporated into the ROSETTA package. The authors provide all experimental datasets used in the supplementary material that other people might use to compare with other methods. The lab that I am working in has a similar software package called Fold-X.

Assuming that the structural coverage of biological parts will continue the current growing trend these structure based methods will become even more useful since one can in principle apply them by modeling the domain of interest by homology.