340 likes | 503 Views
Outline. Background and Related WorkData ComponentsThe ModelThe AlgorithmResults
E N D
1. An EM Algorithm for Inferring the Evolution of Eukaryotic Gene Structure
Liran Carmel, Igor B. Rogozin, Yuri I. Wolf and Eugene V. Koonin
NCBI, NLM, National Institutes of Health
2. Outline Background and Related Work
Data Components
The Model
The Algorithm
Results – Homogeneous Evolution
Results – Heterogeneous Evolution
Summary
3. What are Exons and Introns
4. Related work
5. Outline Background and Related Work
Data Components
The Model
The Algorithm
Results – Homogeneous Evolution
Results – Heterogeneous Evolution
Summary
6. Phylogenetic tree
7. Multiple alignment
8. Presence/absence maps (proteasome component C3)
9. Missing data
10. Missing data (proteasome component C3)
11. Bayesian Network
12. Outline Background and Related Work
Data Components
The Model
The Algorithm
Results – Homogeneous Evolution
Results – Heterogeneous Evolution
Summary
13. Probability structure
14. Rate variation across sites gain variation
loss variation
15. Parameter Summary Global parameters
– probability for intron absence in the root
– fraction of invariant sites
– shape parameters of the gamma distribution
Gene-specific parameters
– gain rate
– loss rate
Branch-specific parameters
– gain coefficient
– loss coefficient
16. Homogeneous vs. Heterogeneous Evolution The number of parameters in the model
17. Outline Background and Related Work
Data Components
The Model
The Algorithm
Results – Homogeneous Evolution
Results – Heterogeneous Evolution
Summary
18. Likelihood maximization via Expectation Maximization E-Step
inward-outward recursions on the tree
member in the junction-tree algorithms family
missing data are naturally embedded
19. Inward (gamma) recursion
20. Inward (gamma) recursion - Initialization
21. Inward (gamma) recursion - Recursion
22. Outward (alpha) recursion
23. Likelihood maximization via EM E-Step
inward-outward recursions on the tree
member in the junction-tree algorithms family
missing data are naturally embedded
M-Step
low-tolerance variable-by-variable maximization
Newton-Raphson
24. Outline Background and Related Work
Data Components
The Model
The Algorithm
Results – Homogeneous Evolution
Results – Heterogeneous Evolution
Summary
25. Intron density in ancient eukaryotes
26. Evolutionary Landscape
27. Modes of Evolution
28. Modes of Evolution
29. Outline
30. Gene Characteristics
31. Combined Features
32. Combined Features
33. Important genes gain introns
34. Outline
35. Conclusions Disparate landscape – both gain and loss play role in intron evolution
The common ancestor of the crown group had an intron content comparable to fungi, apicomlexans and dipterans
Three modes of evolution – more than one mechanism?
Important genes tend to gain introns