MIT Computational Biology: Genomes, Networks, Evolution, Health
Prof. Manolis Kellis
http://compbio.mit.edu/6.047/
Fall 2018

Lecture 04 - HMMs I - Modeling Biological Sequences using Hidden Markov Models

1. Modeling sequential data
- Recognize a type of sequence, genomic, oral, verbal, visual, etc…
2. Definitions
- Markov Chains
- Hidden Markov Models (HMMs)
3. Examples of HMMs
- Recognizing GC-rich regions, preferentially-conserved elements, coding exons, protein-coding gene structures, chromatin states
4. Our first computations
- Running the model: know model, generate sequence of a type
- Evaluation: know model, emissions, states, infer p
- Viterbi: know model, emissions, find optimal path
- Forward: know model, emissions, infer total p over all paths
5. Next time:
- Posterior decoding
- Supervised learning
- Unsupervised learning: Baum-Welch, Viterbi training

Slides for Lecture 4:
https://stellar.mit.edu/S/course/6/fa...