hidden markov modelshisto.ucsf.edu/bms270/bms270a_2012/slides/slides05_hmm.pdf · hidden markov...
TRANSCRIPT
![Page 1: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/1.jpg)
Hidden Markov Models
Mark Voorhies
4/2/2012
Mark Voorhies Hidden Markov Models
![Page 2: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/2.jpg)
Searching with PSI-BLAST
Mark Voorhies Hidden Markov Models
![Page 3: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/3.jpg)
0th order Markov Model
Mark Voorhies Hidden Markov Models
![Page 4: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/4.jpg)
1st order Markov Model
Mark Voorhies Hidden Markov Models
![Page 5: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/5.jpg)
1st order Markov Model
Mark Voorhies Hidden Markov Models
![Page 6: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/6.jpg)
1st order Markov Model
Mark Voorhies Hidden Markov Models
![Page 7: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/7.jpg)
What are Markov Models good for?
Background sequence composition
Spam
Mark Voorhies Hidden Markov Models
![Page 8: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/8.jpg)
Hidden Markov Models
Mark Voorhies Hidden Markov Models
![Page 9: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/9.jpg)
Hidden Markov Models
Mark Voorhies Hidden Markov Models
![Page 10: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/10.jpg)
Hidden Markov Models
Mark Voorhies Hidden Markov Models
![Page 11: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/11.jpg)
Hidden Markov Models
Mark Voorhies Hidden Markov Models
![Page 12: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/12.jpg)
Hidden Markov Models
Mark Voorhies Hidden Markov Models
![Page 13: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/13.jpg)
Hidden Markov Model
Mark Voorhies Hidden Markov Models
![Page 14: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/14.jpg)
The Viterbi algorithm: Alignment
Mark Voorhies Hidden Markov Models
![Page 15: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/15.jpg)
The Viterbi algorithm: Alignment
Dynamic programming, likeSmith-Waterman
Sums best log probabilitiesof emissions and transitions(i.e., multiplyingindependent probabilities)
Result is most likelyannotation of the targetwith hidden states
Mark Voorhies Hidden Markov Models
![Page 16: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/16.jpg)
The Forward algorithm: Net probability
Probability-weighted sumover all possible paths
Simple modification ofViterbi (although summingprobabilities means we haveto be more careful aboutrounding error)
Result is the probability thatthe observed sequence isexplained by the model
In practice, this probabilityis compared to that of a nullmodel (e.g., randomgenomic sequence)
Mark Voorhies Hidden Markov Models
![Page 17: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/17.jpg)
Training an HMM
If we have a set of sequenceswith known hidden states(e.g., from experiment),then we can calculate theemission and transitionprobabilities directly
Otherwise, they can beiteratively fit to a set ofunlabeled sequences that areknown to be true matchesto the model
The most common fittingprocedure is theBaum-Welch algorithm, aspecial case of expectationmaximization (EM)
Mark Voorhies Hidden Markov Models
![Page 18: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/18.jpg)
Training an HMM
If we have a set of sequenceswith known hidden states(e.g., from experiment),then we can calculate theemission and transitionprobabilities directly
Otherwise, they can beiteratively fit to a set ofunlabeled sequences that areknown to be true matchesto the model
The most common fittingprocedure is theBaum-Welch algorithm, aspecial case of expectationmaximization (EM)
Mark Voorhies Hidden Markov Models
![Page 19: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/19.jpg)
Training an HMM
If we have a set of sequenceswith known hidden states(e.g., from experiment),then we can calculate theemission and transitionprobabilities directly
Otherwise, they can beiteratively fit to a set ofunlabeled sequences that areknown to be true matchesto the model
The most common fittingprocedure is theBaum-Welch algorithm, aspecial case of expectationmaximization (EM)
Mark Voorhies Hidden Markov Models
![Page 20: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/20.jpg)
Profile Alignments: Plan 7
(Image from Sean Eddy, PLoS Comp. Biol. 4:e1000069)
Mark Voorhies Hidden Markov Models
![Page 21: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/21.jpg)
Profile Alignments: Plan 7 (from Outer Space)
(Image from Sean Eddy, PLoS Comp. Biol. 4:e1000069)
Mark Voorhies Hidden Markov Models
![Page 22: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/22.jpg)
Rigging Plan 7 for Multi-Hit Alignment
(Image from Sean Eddy, PLoS Comp. Biol. 4:e1000069)
Mark Voorhies Hidden Markov Models
![Page 23: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/23.jpg)
HMMer3 speeds
Eddy, PLoSCompBiol 7:e1002195
Mark Voorhies Hidden Markov Models
![Page 24: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/24.jpg)
HMMer3 sensitivity and specificity
Eddy, PLoSCompBiol 7:e1002195
Mark Voorhies Hidden Markov Models
![Page 25: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/25.jpg)
Homework
Compare the performance of BLASTP, PSI-BLAST, phmmer,and jackhmmer on a difficult sequence such as AGA1p(CAA96325.1). Use the shuffling tool on the course websiteto generate negative controls with the same composition. Forpositive controls, see Euk. Cell 5:628.
Download Cluster3 and JavaTreeView
Read PNAS 95:14863
Mark Voorhies Hidden Markov Models
![Page 26: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/26.jpg)
Stochastic Context Free Grammars
Can emit from both sides → base pairs
Can duplicate emitter → bifurcations
Mark Voorhies Hidden Markov Models
![Page 27: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/27.jpg)
INFERNAL/Rfam
Modified from the INFERNAL User Guide – Nawrocki, Kolbe, and Eddy
Mark Voorhies Hidden Markov Models
![Page 28: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/28.jpg)
INFERNAL/Rfam
Modified from the INFERNAL User Guide – Nawrocki, Kolbe, and Eddy
Mark Voorhies Hidden Markov Models
![Page 29: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/29.jpg)
INFERNAL/Rfam
Modified from the INFERNAL User Guide – Nawrocki, Kolbe, and Eddy
Mark Voorhies Hidden Markov Models
![Page 30: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/30.jpg)
INFERNAL/Rfam
Modified from the INFERNAL User Guide – Nawrocki, Kolbe, and Eddy
Mark Voorhies Hidden Markov Models
![Page 31: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/31.jpg)
INFERNAL/Rfam
Modified from the INFERNAL User Guide – Nawrocki, Kolbe, and Eddy
Mark Voorhies Hidden Markov Models
![Page 32: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/32.jpg)
INFERNAL/Rfam
Modified from the INFERNAL User Guide – Nawrocki, Kolbe, and Eddy
Mark Voorhies Hidden Markov Models
![Page 33: Hidden Markov Modelshisto.ucsf.edu/BMS270/BMS270a_2012/slides/Slides05_HMM.pdf · Hidden Markov Models Mark Voorhies 4/2/2012 Mark Voorhies Hidden Markov Models. Searching with PSI-BLAST](https://reader033.vdocuments.us/reader033/viewer/2022051511/600afc1043245960c377402a/html5/thumbnails/33.jpg)
INFERNAL/Rfam
Modified from the INFERNAL User Guide – Nawrocki, Kolbe, and Eddy
Mark Voorhies Hidden Markov Models