shRNAPrediction

Collaboration between Lowe Lab and Leslie Lab with input from Gunnar Raetsch People working:
 * SHRNA PREDICTION**
 * 1) Vishal Thapar (Lowe Lab)
 * 2) Raphael Pelossof (Leslie Lab)

Ideas to test out for the shRNA prediction
 * 1) Come up with a training set
 * 2) Use the 3 stage data to come up with a better set of candidates for training. The final scoring of the training will be the same as it is now for the selected candidates
 * 3) Which methods to test on this training set?
 * 4) SVM
 * 5) Regression
 * 6) Random forests


 * Meeting Date: 3/19/2012 **
 * Participants: Christian Widmer, Christina Leslie, Christof Fellman, Gunnar Ratsch, Raphael Pelossof, Vishal Thapar **


 * Key takeaways **
 * 1) ** Christof will try to look for and provide the following to the team at MSKCC **
 * 2) ** S5 data **
 * 3) ** Time Between Experiments **
 * 4) ** Gate Points, Numbers or Reference population **
 * 5) ** Right and Left Cut on distribution **
 * 6) ** Individual shRNA data **
 * 7) ** Chicken cells for RNA-Seq **
 * 8) ** Rna-Seq needs to be done for Chicken. That is where the assay was run. **
 * 9) ** We need to do a small/long RNA-Seq experiment on Chicken Cells from Christof **
 * 10) ** Incorporate Secondary Structure into the SVM **
 * 11) ** Hannon Lab is already in the process of validation of their algorithm in different genome wide validations. Their algorithm may not be optimal. Our motivation is to come up with the most optimal algorithm for shRNA prediction **
 * 12) ** (Please add more) **


 * Next Steps in the project: **


 * 1) ** Vishal: Run Random Forest classification on the current data to replicate Simons algorithm. Deadline: April 5th. **
 * 2) ** Raphael: **
 * 3) ** Christian: **

Things to accomplish Week of 4/2/2012
 * 1) For Random Forest: Run the random forest on original data sent by R
 * 2) For Lasso prediction:
 * 3) Run a subset for lasso on original data and calculate the ROC.

Data

Algorithms