Download presentation
Presentation is loading. Please wait.
1
Sequence motifs, information content, logos, and Weight matrices
Morten Nielsen, CBS, BioCentrum, DTU
2
Objectives Visualization of binding motifs
Construction of sequence logos Understand the concepts of weight matrix construction One of the most important methods of bioinformatics
3
Outline Pattern recognition Information content
Regular expressions and probabilities Information content Sequence logos Kullback-Leibler Multiple alignment and sequence motifs Weight matrix construction Sequence weighting Low (pseudo) counts Calculate a weight matrix from a sequence alignment Examples from the real world MHC class II binding Gibbs sampling
4
Host/virus interaction. Sars infection
5
Immune system
6
Processing of intracellular proteins
7
Encounter with death
8
Binding Motif. MHC class I with peptide
Anchor positions
9
Sequence information SLLPAIVEL YLLPAIVHI TLWVDPYEV GLVPFLVSV KLLEPVLLL LLDVPTAAV LLDVPTAAV LLDVPTAAV LLDVPTAAV VLFRGGPRG MVDGTLLLL YMNGTMSQV MLLSVPLLL SLLGLLVEV ALLPPINIL TLIKIQHTL HLIDYLVTS ILAPPVVKL ALFPQLVIL GILGFVFTL STNRQSGRQ GLDVLTAKV RILGAVAKV QVCERIPTI ILFGHENRV ILMEHIHKL ILDQKINEV SLAGGIIGV LLIENVASL FLLWATAEA SLPDFGISY KKREEAPSL LERPGGNEI ALSNLEVKL ALNELLQHV DLERKVESL FLGENISNF ALSDHHIYL GLSEFTEYL STAPPAHGV PLDGEYFTL GVLVGVALI RTLDKVLEV HLSTAFARV RLDSYVRSL YMNGTMSQV GILGFVFTL ILKEPVHGV ILGFVFTLT LLFGYPVYV GLSPTVWLS WLSLLVPFV FLPSDFFPS CLGGLLTMV FIAGNSAYE KLGEFYNQM KLVALGINA DLMGYIPLV RLVTLKDIV MLLAVLYCL AAGIGILTV YLEPGPVTA LLDGTATLR ITDQVPFSV KTWGQYWQV TITDQVPFS AFHHVAREL YLNKIQNSL MMRKLAILS AIMDKNIIL IMDKNIILK SMVGNWAKV SLLAPGAKQ KIFGSLAFL ELVSEFSRM KLTPLCVTL VLYRYGSFS YIGEVLVSV CINGVCWTV VMNILLQYV ILTVILGVL KVLEYVIKV FLWGPRALV GLSRYVARL FLLTRILTI HLGNVKYLV GIAGGLALL GLQDCTMLV TGAPVTYST VIYQYMDDL VLPDVFIRC VLPDVFIRC AVGIGIAVV LVVLGLLAV ALGLGLLPV GIGIGVLAA GAGIGVAVL IAGIGILAI LIVIGILIL LAGIGLIAA VDGIGILTI GAGIGVLTA AAGIGIIQI QAGIGILLA KARDPHSGH KACDPHSGH ACDPHSGHF SLYNTVATL RGPGRAFVT NLVPMVATV GLHCYEQLV PLKQHFQIV AVFDRKSDA LLDFVRFMG VLVKSPNHV GLAPPQHLI LLGRNSFEV PLTFGWCYK VLEWRFDSR TLNAWVKVV GLCTLVAML FIDSYICQV IISAVVGIL VMAGVGSPY LLWTLVVLL SVRDRLARL LLMDCSGSI CLTSTVQLV VLHDDLLEA LMWITQCFL SLLMWITQC QLSLLMWIT LLGATCMFV RLTRFLSRV YMDGTMSQV FLTPKKLQC ISNDVCAQV VKTDGNPPE SVYDFFVWL FLYGALLLA VLFSSDFRI LMWAKIGPV SLLLELEEV SLSRFSWGA YTAFTIPSI RLMKQDFSV RLPRIFCSC FLWGPRAYA RLLQETELV SLFEGIDFY SLDQSVVEL RLNMFTPYI NMFTPYIGV LMIIPLINV TLFIGSHVV SLVIVTTFV VLQWASLAV ILAKFLHWL STAPPHVNV LLLLTVLTV VVLGVVFGI ILHNGAYSL MIMVKCWMI MLGTHTMEV MLGTHTMEV SLADTNSLA LLWAARPRL GVALQTMKQ GLYDGMEHL KMVELVHFL YLQLVFGIE MLMAQEALA LMAQEALAF VYDGREHTV YLSGANLNL RMFPNAPYL EAAGIGILT TLDSQVMSL STPPPGTRV KVAELVHFL IMIGVLVGV ALCRWGLLL LLFAGVQCQ VLLCESTAV YLSTAFARV YLLEMLWRL SLDDYNHLV RTLDKVLEV GLPVEYLQV KLIANNTRV FIYAGSLSA KLVANNTRL FLDEFMEGV ALQPGTALL VLDGLDVLL SLYSFPEPE ALYVDSLFF SLLQHLIGL ELTLGEFLK MINAYLDKL AAGIGILTV FLPSDFFPS SVRDRLARL SLREWLLRI LLSAWILTA AAGIGILTV AVPDEIPPL FAYDGKDYI AAGIGILTV FLPSDFFPS AAGIGILTV FLPSDFFPS AAGIGILTV FLWGPRALV ETVSEQSNV ITLWQRPLV
10
Sequence Information Say that a peptide must have L at P2 in order to bind, and that A,F,W,and Y are found at P1. Which position has most information? How many questions do I need to ask to tell if a peptide binds looking at only P1 or P2? P1: 4 questions (at most) P2: 1 question (L or not) P2 has the most information Calculate pa at each position Entropy Information content Conserved positions PV=1, P!v=0 => S=0, I=log(20) Mutable positions Paa=1/20 => S=log(20), I=0
11
Information content A R N D C Q E G H I L K M F P S T W Y V S I
12
Sequence logos Height of a column equal to I
HLA-A0201 Height of a column equal to I Relative height of a letter is p Highly useful tool to visualize sequence motifs High information positions
13
Kullback Leibler Logo
14
Characterizing a binding motif from small data sets
10 MHC restricted peptides What can we learn? A at P1 favors binding? I is not allowed at P9? K at P4 favors binding? Which positions are important for binding? ALAKAAAAM ALAKAAAAN ALAKAAAAR ALAKAAAAT ALAKAAAAV GMNERPILT GILGFVFTM TLNAWVKVV KLNEPVLLL AVVPFIVSV
15
Simple motifs Yes/No rules
10 MHC restricted peptides ALAKAAAAM ALAKAAAAN ALAKAAAAR ALAKAAAAT ALAKAAAAV GMNERPILT GILGFVFTM TLNAWVKVV KLNEPVLLL AVVPFIVSV Only 11 of 212 peptides identified! Need more flexible rules If not fit P1 but fit P2 then ok Not all positions are equally important We know that P2 and P9 determines binding more than other positions Cannot discriminate between good and very good binders
16
Simple motifs Yes/No rules
10 MHC restricted peptides ALAKAAAAM ALAKAAAAN ALAKAAAAR ALAKAAAAT ALAKAAAAV GMNERPILT GILGFVFTM TLNAWVKVV KLNEPVLLL AVVPFIVSV Example Two first peptides will not fit the motif. They are all good binders (aff< 500nM) RLLDDTPEV 84 nM GLLGNVSTV 23 nM ALAKAAAAL 309 nM
17
Extended motifs Fitness of aa at each position given by P(aa)
Example P1 PA = 6/10 PG = 2/10 PT = PK = 1/10 PC = PD = …PV = 0 Problems Few data Data redundancy/duplication ALAKAAAAM ALAKAAAAN ALAKAAAAR ALAKAAAAT ALAKAAAAV GMNERPILT GILGFVFTM TLNAWVKVV KLNEPVLLL AVVPFIVSV RLLDDTPEV 84 nM GLLGNVSTV 23 nM ALAKAAAAL 309 nM
18
Sequence information Raw sequence counting
ALAKAAAAM ALAKAAAAN ALAKAAAAR ALAKAAAAT ALAKAAAAV GMNERPILT GILGFVFTM TLNAWVKVV KLNEPVLLL AVVPFIVSV
19
} Sequence weighting Poor or biased sampling of sequence space
ALAKAAAAM ALAKAAAAN ALAKAAAAR ALAKAAAAT ALAKAAAAV GMNERPILT GILGFVFTM TLNAWVKVV KLNEPVLLL AVVPFIVSV } Similar sequences Weight 1/5 Poor or biased sampling of sequence space Example P1 PA = 2/6 PG = 2/6 PT = PK = 1/6 PC = PD = …PV = 0 RLLDDTPEV 84 nM GLLGNVSTV 23 nM ALAKAAAAL 309 nM
20
Sequence weighting ALAKAAAAM ALAKAAAAN ALAKAAAAR ALAKAAAAT ALAKAAAAV
GMNERPILT GILGFVFTM TLNAWVKVV KLNEPVLLL AVVPFIVSV
21
Pseudo counts ALAKAAAAM
ALAKAAAAN ALAKAAAAR ALAKAAAAT ALAKAAAAV GMNERPILT GILGFVFTM TLNAWVKVV KLNEPVLLL AVVPFIVSV I is not found at position P9. Does this mean that I is forbidden (P(I)=0)? No! Use Blosum substitution matrix to estimate pseudo frequency of I at P9
22
What is a pseudo count? Say V is observed at P1
A R N D C Q E G H I L K M F P S T W Y V A R N D C …. Y V Say V is observed at P1 Knowing that V at P1 binds, what is the probability that a peptide could have I at P1? P(I|V) = 0.16
23
Pseudo count estimation
ALAKAAAAM ALAKAAAAN ALAKAAAAR ALAKAAAAT ALAKAAAAV GMNERPILT GILGFVFTM TLNAWVKVV KLNEPVLLL AVVPFIVSV Calculate observed amino acids frequencies fa Pseudo frequency for amino acid b Example
24
Weight on pseudo count ALAKAAAAM ALAKAAAAN ALAKAAAAR ALAKAAAAT ALAKAAAAV GMNERPILT GILGFVFTM TLNAWVKVV KLNEPVLLL AVVPFIVSV Pseudo counts are important when only limited data is available With large data sets only “true” observation should count is the effective number of sequences (N-1), is the weight on prior
25
Weight on pseudo count Example
ALAKAAAAM ALAKAAAAN ALAKAAAAR ALAKAAAAT ALAKAAAAV GMNERPILT GILGFVFTM TLNAWVKVV KLNEPVLLL AVVPFIVSV Example If large, p ≈ f and only the observed data defines the motif If small, p ≈ g and the pseudo counts (or prior) defines the motif is [50-200] normally
26
Sequence weighting and pseudo counts
ALAKAAAAM ALAKAAAAN ALAKAAAAR ALAKAAAAT ALAKAAAAV GMNERPILT GILGFVFTM TLNAWVKVV KLNEPVLLL AVVPFIVSV RLLDDTPEV 84nM GLLGNVSTV 23nM ALAKAAAAL 309nM P7P and P7S > 0
27
Position specific weighting
We know that positions 2 and 9 are anchor positions for most MHC binding motifs Increase weight on high information positions Motif found on large data set
28
Weight matrices Estimate amino acid frequencies from alignment including sequence weighting and pseudo count What do the numbers mean? P2(V)>P2(M). Does this mean that V enables binding more than M. In nature not all amino acids are found equally often In nature V is found more often than M, so we must somehow rescale with the background qM = 0.025, qV = 0.073 Finding 5% V is hence not significant, but 5% M highly significant A R N D C Q E G H I L K M F P S T W Y V
29
Weight matrices A weight matrix is given as Wij = log(pij/qj)
where i is a position in the motif, and j an amino acid. qj is the background frequency for amino acid j. W is a L x 20 matrix, L is motif length A R N D C Q E G H I L K M F P S T W Y V
30
Scoring a sequence to a weight matrix
Score sequences to weight matrix by looking up and adding L values from the matrix A R N D C Q E G H I L K M F P S T W Y V Which peptide is most likely to bind? Which peptide second? RLLDDTPEV GLLGNVSTV ALAKAAAAL 11.9 14.7 4.3 84nM 23nM 309nM
31
An example!! (See handout)
33
The Blosum matrix A R N D C Q E G H I L K M F P S T W Y V
35
Example from real life 10 peptides from MHCpep database
Bind to the MHC complex Relevant for immune system recognition Estimate sequence motif and weight matrix Evaluate motif “correctness” on 528 peptides ALAKAAAAM ALAKAAAAN ALAKAAAAR ALAKAAAAT ALAKAAAAV GMNERPILT GILGFVFTM TLNAWVKVV KLNEPVLLL AVVPFIVSV
36
Prediction accuracy Measured affinity Prediction score
Pearson correlation 0.45 Measured affinity Prediction score
37
Predictive performance
38
Class II MHC binding MHC class II binds peptides in the class II antigen presentation pathway Binds peptides of length 9-18 (even whole proteins can bind!) Binding cleft is open Binding core is 9 aa
39
Gibbs sampler www.cbs.dtu.dk/biotools/EasyGibbs
100 10mer peptides 2100~1030 combinations Monte Carlo simulations can do it
40
Gibbs sampler. Prediction accuracy
41
Gibbs sampling can detect MHC class II binding motif
Summary Weight matrices can accurately describe a sequence motif like MHC class I Use sequence weighting to remove data redundancy Use pseudo count to compensate for few data points Gibbs sampling can detect MHC class II binding motif
Similar presentations
