Reading questions

1. Why do we need a probabilistic model for a TF binding site?

2.  What is the difference between PSSM and FMM?  What are pros and cons of FMM?

3. Describe the proposed computational approach: probabilistic model; training data: observed/ hidden variables; evaluation methods.

4. Is there any room for improvement, either in terms of modeling or learning?

(optional) 5. Give examples of other problems where replacing a PSSM-type model with FMM-type can improve the performance.