Transcript Slide

The Immune Response
The humoral response
involves interaction of B
cells with antigen (Ag) and
their differentiation into
antibody-secreting plasma
cells. The secreted antibody
(Ab) binds to the antigen
and facilitates its clearance
from the body.
The cell-mediated
responses involve various
subpopulations of T cells
that recognize antigen
presented on self-cells.
Helper T cells respond to
antigen by producing
cytokines. Cytotoxic T cells
respond to antigen by
developing into cytotoxic T
lymphocytes (CTLs), which
mediate killing of altered
self-cells (e.g., virusinfected cells).
The MHC class I pathway
Antigen
Not all peptides binding to
MHC molecules are epitopes,
but all T-cell epitopes need to
bind to MHC.
Proteasome
Identifying of T-cell epitopes is
important for development of
peptide-based vaccines,
evaluation of subunit
vaccines, diagnostic
development
Peptides
TAP
T-cell epitope
ER
MHC I
TCD8+
Antigen Presenting Cell
T cell receptor (TCR)
Cytotoxic T
lymphocyte
(CTL)
Xenoreactive Complex AHIII 12.2 TCR bound
to P1049 (ALWGFFPVLS) /HLA-A2.1
T cell epitope – a short linear
peptide or other chemical entity
(native or denatured antigen) that
binds MHC (class I binds 8-10 aa
peptides; class II binds 11-25 aa
peptides) and may be recognized
by T-cell receptor (TCR).
T-Cell
Receptor
V
V
Epitope is a peptide which able to elicit
T cell response.
T cell recognition of antigen involves
tertiary complex “antigen-TCRMHC”.
MHC
class I
-2-Microglobulin
1lp9
MHC class I facts
MHC class I in human is called HLA I (Human Leukocyte Antigen) (in mouse H-2).
Every normal (heterozygous) human expresses six different MHC class I
molecules on every cell, containing α-chains derived from the two alleles of HLA-A,
HLA-B, HLA-C genes that inherited from the parents.
MHC genes are the most polymorphic in human genome. For each locus hundreds
of different alleles exist. For today, there are known 489 HLA-A alleles, 830 HLA-B
and 266 HLA-C alleles (1,670 alleles including non-classical 7 alleles). Some of
these alleles are more closely related to the alleles found in chimpanzees than to
another human alleles from the same gene.
The role of MHC diversity in sexual selection: the more diverse the MHC genes of
parents, the stronger the immune system of the offspring. A preference of mate of
different MHC was demonstrated on both mice and humans.
The MHC molecules of an individual do not discriminate between foreign and self
peptides.
Prediction of MHC class I binding peptides
Regression models predicting the peptide-MHC binding affinity.
Require a lot of experimental data pairs pairs {peptide – affinity
value} for a given MHC allele.
Such data are currently available for very restricted number of
alleles (<50 for HLA class I).
Sequence-based methods.
Classification models distinguishing binders from non-binders.
Do not require consistent quantitative binding data.
QSAR, 3D-QSAR approach
Average relative binding matrices (sequence-based approach)
Structure-based methods: docking, threading (slow - atom
energy function calculation, fast – knowledge-based residue
contact scoring functions) etc.
Performance measures for prediction methods
ROC curve
TP
FP
threshold
FN
TN
sensitivity = TP / (TP + FN) =
6/7= 0.86
specificity = TN / (TN + FP) =
6/8 = 0.75
True positive rate, TP / (TP +
FN)
1
0.9
0.8
0.7
0.6
0.5
AROC
0.4
0.3
0.2
0.1
0
0
0.2
0.4
0.6
0.8
False positive rate, FP / (FP + TN)
1
Sequence-based methods for prediction of peptide
binding to MHC class I
ALAKAAAAM
ALAKAAAAN
ALAKAAAAV
ALAKAAAAT
GMNERPILT
GILGFVFTM
TLNAWVKVV
Gibbs sampling
Sequence motifs, matrices
Sequence weighted matrices
Hidden Markov Models
Artificial Neural Networks
SVM
0.95
0.9
KLNEPVLLL
AVVPFIVSV
Peptides
known to bind
to the HLAA*0201
molecule.
Performance (measured
as AROC) depends on the
number of training
peptides
Aroc
0.85
0.8
0.75
0.7
0.65
2
10
20
100
200
Number of training peptides
500
Benchmarking predictions of peptide binding to MHC I
(Peters et al. PLoS Comput Biol. 2006 Jun 9;2(6):e65)
Data: pairs {peptide – affinity value in terms of IC50 nM}
for a given MHC allele
48 different mouse, human (35 HLA class I), macaque,
and chimpanzee MHC class I alleles.
Length of peptides 8 – 11 aa.
48,828 experimental data points.
20 different methods were evaluated.
Peptide binding to MHC class I affinity prediction methods comparison
Correlation coefficients (ARB=0.55, SMM=0.62, ANN=0.69) are significantly different (p<0.05
using a t test).
Aroc values (ARB=0.934, SMM=0.952, ANN=0.957) are significantly different (p<0.05 using a
paired t test on Aroc values generated by bootstrap).
Predep – structure-based method of Schueler-Furman, Altuvia, Sette, Margalit
(2000) Protein Sci.
Structure-based methods for prediction of
peptide-MHC I binding
Contrarily to sequence-based methods structure-based
methods are applicable to different peptide length and
MHC alleles.
Sequence-based methods have limited structural
interpretability contrarily to structure-based methods.
More than 40 X-ray structures of different peptide-MHC I
complexes are available (only 10 different HLA class I
allotypes).
Learning MHC I – peptide binding
Data:
HLA I sequences
Input: 37 3D-structures of MHC-peptide complexes
MHC-binding affinity data: 870 data points (quantitative data)
Binders and non-binders (peptides) from 3 databases known for particular
HLA alleles (binary data)
Method: threading of a peptide sequence onto 3D-structure of a complex of other
peptide with the same or similar (by sequence) HLA molecule combined with
machine learning on binding data
binding energy is additive,
the residue pairwise potentials depend only on the amino acids (not on
their context),
20x20 matrix of pairwise potentials is derived from known 3D-structures
and binding data
Parameters of binding energy function are learned on binding data
Results:
Proposed method outperforms original threading method in case when
both structure and binding data are available for the allele
Proposed method performs similar to the threading when binding data are
used for similar alleles; while adding of binary data known for the allele
improves the prediction
HLA
Number of
Peptide
s
AUC ADT
AUC ANN
Best Online
Tool
A_0101
1158
0.9657
0.9798
0.955
hla ligand
A_0201
3090
0.9521
0.9564
0.922
hla_a2_smm
A_0202
1448
0.9033
0.8988
0.793
multipredann
A_0203
1444
0.9141
0.9203
0.788
multipredann
A_0206
1438
0.9191
0.9261
0.735
multipredann
A_0301
2095
0.9298
0.9366
0.851
multipredann
A_1101
1986
0.9442
0.9511
0.869
multipredann
A_2301
105
0.8044
0.8514
A_2402
198
0.7852
0.822
A_2403
255
0.8784
0.9175
A_2601
673
0.9224
A_2902
161
A_3001
multipredann
0.77
syfphethi
0.9552
0.736
pepdist
0.8866
0.9317
0.597
rankpep
670
0.941
0.945
A_3002
93
0.7633
0.744
A_3101
1870
0.9313
0.9274
0.829
bimas
A_3301
1141
0.9363
0.9141
0.807
pepdist
A_6801
1142
0.8847
0.8823
0.772
syfphethi
A_6802
1435
0.8963
0.8986
0.643
mhcpred
A_6901
834
0.8902
0.8803
HLA
Number of
Peptide
s
AUC ADT
AUC ANN
Best Online
Tool
B_0702
1263
0.9573
0.9636
0.942
hlaligand
B_0801
709
0.854
0.9533
0.766
pepdist
B_1501
979
0.9075
0.942
0.816
rankpep
B_1801
119
0.8687
0.838
0.779
pepdist
B_2705
970
0.9217
0.9371
0.926
bimas
B_3501
737
0.8691
0.8739
0.792
bimas
B_4001
1079
0.8933
0.9155
B_4002
119
0.8186
0.7524
0.775
rankpep
B_4402
120
0.6775
0.7785
0.783
syfphethi
B_4403
120
0.6239
0.7634
0.628
rankpep
B_4501
115
0.8015
0.8609
B_5101
245
0.8474
0.8856
0.82
pepdist
B_5301
255
0.8934
0.8974
0.861
rankpep
B_5401
256
0.8457
0.9025
0.799
svmhc
B_5701
60
0.832
0.8246
0.767
pepdist
B_5801
989
0.94
0.96
0.899
bimas
ANN method (Nielsen et al. 2003) outperformed all other sequence-based
methods. The proposed methods can outperform ANN when the available
training data for an allele is small.
HIV virus evolves to modulate its binding to MHC molecules
Hypothesis: HIV mutations correlate with the MHC types of the host
in a way that the virus whose peptides bind well to a particular MHC
molecule is typically under strong immune pressure in patients with
this MHC type, and it is forced to mutate away (escape) from its
fittest form towards a form that binds less well to MHC.
Data: 246 HIV patients, >1,000 HIV sequences
Results for the most frequent (in the Western world) MHC allele
A0201: significant negative correlation between HIV peptide-MHC
binding energy calculated values and viral load reflecting HIV
mutations towards escaping of the immune response (lower MHC
binding).
The concept of MHC sypertypes
MHC polymorphism is essential to protect the population from
invasion by pathogens. But it generates problem for epitope-based
vaccine design: a vaccine needs to contain a unique epitope binding
each MHC allele.
A factor that may reduce the number of epitopes necessary to
include in a vaccine is that many of different HLA molecules have
similar specificity, i.e. bind similar by sequence peptides. Such HLAs
represents a supertype.
HLA-A were classified on 5 supertypes (with a number of nonclassified alleles):
A2 – hydrophobic amino acid in binding peptide position 9
A3 – basic aa in position 9
A26 – acidic aa in position 1
A1 – acidic aa in position 3
A24 – tyrosine in position 2