Attendance: Ranjit Randhawa, John Archie, Naren Ramakrishnan, Ceci Vasquez-Robinet, Xiaofeng Bao.
Review of where people are:
John A is looking at TonE elements, aims to generate graphs of proximity to gene versus frequency of element.
Dr Grene wants to look upstream for response elements (RE) and look at combinations of RE, she wants to study the arrangement of these elements. (Ranjit)
Overall goal from Dr R.:
- We would like to look at every gene, and look upstream of that gene.
- Want to see combinations of all REs that occur.
- Want distance for each one.
- Bao will take distance of each and model some context.
Gaussian Processes: Used to model.
E.g. GPA and height of students
Expect taller person to have higher GPA. Think of the function as ?smooth.? The idea is to relate x1-x2 distance to y1-y2 distance. Covariance between the two y points is the same function as covariance between the two x points.
Idea in Gaussian Processes is that things next to each other are similar.
With RE we want to look at things that are proximal, are they doing something similar? If so what is the notion of difference.
Covariance structure of constellation of RE
- Trying to relate how REs combine together and what they create
- Want to use this on different species.
We want to express sequence in terms of classes. Then mine covariant structure along these lines. We want to also take multiple genes and get their covariant structure.
Ultimately we want to characterize genomic sequences and how they match up together -> what is their constellation
E.g. Potential use
- Look at gene that has only RE1 (A)
- Look at gene that has only RE2 (B)
- Look at gene that has both RE1 and RE2 (C)
- Find covariant structure of A and B
- Is covariant structure C a combination of A and B?
Next meeting Dr R and Bao will talk further about Gaussian Processes.