The SHAPE procedure used in Eterna interacts directly with the backbone, not the A, C, G and U nucleic acids themselves. Although there is a high correlation between SHAPE values at a particular base position and whether or not the the ribonucleic acid at that position is paired with another base (C with G, A with U, or G with U), the correlation is not perfact. I think the SHAPE score is telling us more than just the secondary structure; we just need to figure out how to make use of that information.
General plan of attack
- Write up some intriguing examples, to perhaps get others interested.
- Continue the development of scripts to streamline the analysis
- Analyze data from one or more existing labs, and formulate an initial model for the relationship between backbone conformers and SHAPE scores
- Run one or more new labs to test the model. The test would be whether backbone conformers taken from 3D structure prediction can predict lab SHAPE results better than the simple binary decision (protected or not) based on the secondary structure prediction of base pairings.
- If the above is successful, develop a procedure for using the SHAPE results to improve 3D structure prediction.