Read Microsoft PowerPoint - ImmunoiformatincsHarvard2007.ppt text version

Immunoinformatics: From Molecules to Vaccines

Vladimir Brusic Cancer Vaccine Center Dana-Farber Cancer Institute

Wikipedia: Computational biology: an interdisciplinary field that applies the techniques of computer science and applied mathematics to biology Bioinformatics: applies algorithms and statistical techniques to biological datasets, typically large numbers of DNA, RNA, or protein sequences Immunoinformatics: bioinformatics applied to the study of immune system and its function

V. Brusic 2007

SUMMARY Introduction Immunological databases Analysis of viral sequences Computational models Conclusions and future developments

V. Brusic 2007

The Immune System composed of many interdependent organs and tissues, cell types, and molecules protects the organism from infections (bacterial, parasitic, fungal, or viral) and from the growth of tumor cells; it underpins homeostasis the second most complex system in human body

V. Brusic 2007

ORGANS AND TISSUES OF THE IMMUNE SYSTEM

V. Brusic 2007

CELLS OF THE IMMUNE SYSTEM

Stem cell Lymphoid progenitor Myeloid progenitor

B cell Natural killer cell

T cell Neutrophil Eosinophil Basophil Mast cell

Monocyte

Plasma cell

Memory B cell

Helper T cell

Killer T cell

Dendritic cell

Macrophage

BILLIONS OF CLONES

V. Brusic 2007

MOLECULES OF THE IMMUNE SYSTEM

?

V. Brusic 2007

An enormous diversity in human immune system >1013 MHC class I haplotypes (IMGT-HLA) 1012 different T-cell receptors (Arstila et al., 1999) 1012 B-cell clonotypes in an individual (Jerne, 1993) >109 combinatorial antibodies (Jerne, 1993) 1011 linear epitopes composed of nine amino acids >>1011 conformational epitopes

V. Brusic 2007

Immune system is a complex system Immunology is essentially a combinatorial science Additional information

multi-step processing pathways network-type interactions complex signalling mechanisms for modulation of immune responses

Current data are only a tiny fraction of possible situations and the amount of information keeps growing Deciphering specific mechanisms of immune responses or correcting undesirable immune responses is increasingly dependent on using computational methods

V. Brusic 2007

COMPUTER COMPUTER SCIENCE SCIENCE

IMMUNOLOGY IMMUNOLOGY

Learning Algorithms, Pattern Recognition, Adaptive Memories, Intelligent Agents

IMMUNOINFORMATICS

Design of Experiments, Data Interpretation

DATABASES DATABASES

ANALYTICAL TOOLS

COMPUTATIONAL COMPUTATIONAL EXPERIMENTS EXPERIMENTS

V. Brusic 2007

Immunoinformatics

Database technology for storage, analysis, and modelling of immunological data Sequence analysis and various statistical tools Computational models to facilitate research in immunology - molecular level models - system level models

V. Brusic 2007

Genomics vs. Immunomics Genomics ­ deciphering genetic code 105 genes coding for 106 products Immunomics ­ deciphering immune response 103-104 genes coding for >1012 products Immunomics ­ huge diversity, focus on function and clinical outcomes

V. Brusic 2007

INNOVATION CURVE Forecoming Growth Boom in Immunomics

1 0.9 Genomics 0.8 Immunomics 0.7 INNOVATION 0.6 0.5 MATURITY 0.4 GROWTH BOOM 0.3 0.2 INNOVATION 0.1 0 1985 1995 2005

Adoption rate

2015

2025

Year

Supported by technologies and instrumentation for high-throughput screening

SUMMARY Introduction Immunological databases Analysis of viral sequences Computational models Conclusions and future developments

V. Brusic 2007

Biological databases

General databases

Specialist immunological databases

Data warehouses

V. Brusic 2007

Biological databases

General databases

Large quantities of data Basic annotations Limited analysis tools

Specialist immunological databases

Data warehouses

V. Brusic 2007

Biological databases

General databases

Large quantities of data Basic annotations Limited analysis tools

Specialist immunological databases

Limited quantities of data Extended annotations by experts More extensive analysis tools

Data warehouses

V. Brusic 2007

Biological databases

General databases

Large quantities of data Basic annotations Limited analysis tools

Specialist immunological databases

Limited quantities of data Extended annotations by experts More extensive analysis tools

Data warehouses

Combines data from multiple sources Emphasis on specific analyses

V. Brusic 2007

V. Brusic 2007

Lessons learnt

Databases are expensive to maintain and keep up-to-date Information is scattered across multiple sources Critical information for specific use is typically lacking and needs to be added (e.g. functional classification) Intelligent linking of data sources with analytical tools for data mining offers significant advantages The biological/clinical question must be well-defined

V. Brusic 2007

Example

IUIS Allergen repository Sequence databases PubMed abstracts Cross-reactivity data

Literature

Allergen sequences TOOLS: Search, BLAST, 3D visualisation, allergenicity, allergic cross-reactivity

V. Brusic 2007

ALLERDB data warehouse

Example

IUIS Allergen repository Sequence Literature MANUAL databases Allergen sequences TOOLS: Search, BLAST, 3D visualisation, allergenicity, allergic cross-reactivity

V. Brusic 2007

PubMed abstracts Cross-reactivity data

ALLERDB data warehouse

V. Brusic 2006

Example

IUIS Allergen repository Sequence Literature MANUAL databases Allergen sequences TOOLS: Search, BLAST, 3D visualisation, allergenicity, allergic cross-reactivity

V. Brusic 2007

Cross-reactivity data

TEXT MINING

PubMed abstracts

ALLERDB data warehouse

V. Brusic 2006

Miotto O, Tan TW. Brusic V. (2005). Extraction by example: induction of structural rules for the analysis of molecular sequence data from heterogeneous sources. LNCS 3578, 398-405. Miotto O, Tan TW, Brusic V. (2005). Supporting the curation of biological databases with reusable text mining. Genome Informatics 16(2), 32-44.

Example

IUIS Allergen repository Sequence Literature MANUAL databases Allergen sequences

Cross-reactivity data

TEXT MINING

PubMed abstracts

CLEARLY TOOLS: DEFINED Search, BLAST, 3D visualisation, PURPOSE

V. Brusic 2007

allergenicity, allergic cross-reactivity

ALLERDB data warehouse

V. Brusic 2006

Example

IUIS Allergen repository Sequence Literature MANUAL databases Allergen sequences

Cross-reactivity data

TEXT MINING

PubMed abstracts

CLEARLY TOOLS: DEFINED Search, BLAST, 3D visualisation, PURPOSE

V. Brusic 2007

allergenicity, allergic cross-reactivity

DISCOVERED A NEW CONCEPT OF ALLERGEN ALLERDB NETWORKS data warehouse

V. Brusic 2006

ADVANTAGES OF THE APPROACH Rapid assembly of data and tools Database frozen, but data fully preserved Re-analysis of data can be performed at the time of updating Updates can be performed very fast Additional analyses can be "plugged in" e.g. study of viral diversity for vaccine targets

V. Brusic 2007

SUMMARY Introduction Immunological databases Analysis of viral sequences Computational models Conclusions and future developments

V. Brusic 2007

INFLUENZA VIRUS PARTICLE

(b) NA HA (a) PB1 PA PB2 M1 M2

RNA NP

RNP particle PB2 PB1 PA HA NA NP M NS

a) Influenza A ribonucleoprotein (RNP) particle consists of viral RNA, nucleoprotein (NP) and polymerase complex (PA, PB1, PB2). (b) Structure of influenza A virus

RNA Segment 1 2 3 4 5 6 7a 7b 8a 8b

Protein Product Polymerase basic protein 2 (PB2) Polymerase basic protein 1 (PB1) Polymerase acidic protein (PA) Hemagglutinin (HA) Nucleoprotein (NP) Neuraminidase (NA) Matrix protein 1 (M1) Matrix protein 2 (M2) Nonstructural protein 1 (NS1) Nonstructural protein 2 (NS2)

Protein Length 759 aa 757 aa 716 aa 566 aa 498 aa 454 aa 252 aa 96 aa 230 aa 121 aa

GenPept Accession NP_040987 NP_040985 NP_040986 NP_040980 NP_040982 NP_040981 NP_040978 NP_040979 NP_040984 NP_040983

MULTI-DIMENSIONAL PROBLEM Composition of the viral proteome High viral diversity Variation of human immune responses

(Brusic and August, Pharmacogenomics 2004).

V. Brusic 2007

Analysis of antigenic diversity of viruses

Commonly used computational approach is molecular phylogenetic analysis which is mainly relevant for genetic diversity Antigenic and genetic diversity of viruses correlate at times, but antigenic diversity often results in greater functional differences than expected from genetic differences revealed by molecular phylogenetic analysis (eg. Morvan et al, 1990) We developed a novel approach for systematic antigenic diversity analysis of large sets of viral sequences, based on the analysis of 9-mer peptides Advantages can be applied to complete and partial sequences unlimited number of sequences can be analysed simple metrics can be expanded to non-contiguous sequences

V. Brusic 2007

123456789

MKTKAACSTLLVLLCA...

9-mer

MRTKAACSTLLVLLCA MKTKAACSTLLVLTCA MRTKAACSTLLVLTCA

Antigenically redundant

Protein subtype HA Subtype H1 H2 H3 H5 H7 H9 Total HA NA Subtype N1 N2 N7 N8 Total NA M1 M2 NS1 NS2 PB1 PB2 PA NP Not classifiable TOTAL

Total no. of sequences (all) 622 105 1214 287 255 288 2771 324 686 22 46 1078 779 458 854 573 558 546 512 1048 324

Number of sequences from human isolates Total 423 75 970 26 7 8 1509 133 382 6 1 522 275 214 219 182 182 143 130 467 3843 Complete1 57 17 84 25 4 7 194 110 340 3 1 454 239 206* 205 182 118 119 109 261 2087 Incomplete2 366 58 886 1 3 1 1315 23 42 3 68 36 8 14 64 24 21 206 1756

Data collection Data processing

NCBI Entrez protein database

Extraction of portions corresponding to reference sequence for each protein Remove duplicate sequences (get non-redundant sequences)

Data analysis Antigenic analysis of whole protein sequences Effects of various parameters on antigenic diversity

Antigenic analysis of immunological hotspots

Remove antigenically redundant sequences

Remove 9-mers with single AA difference

Identification of imunological hotspots

Remove antigenically redundant sequences

Remove antigenically redundant peptides

TARGETS

V. Brusic 2007

NUMBER OF SEQUENCES

100

20

LENGTH OF SEQUENCES

18 Number of sequences to cover 100% antigenic diversity 16

90

80 Percentage of antigenic diversity covered (%)

70

10 sequences in test dataset 15 sequences in test dataset

14

60

20 sequences in test dataset

12

50

10

40

8

30

6

20

4

10

2

0 0 2 4 6 8 10 12 14 16 18 Number of sequences to cover 100% antigenic diversity

0 0 100 200 300 400 500 Length of fragments in test dataset

12

REGION OF SEQUENCES Parameters that affect antigenic diversity of viral sequences. This is study is for the West Nile virus envelope protein. Largest effect ­ length of sequence Some effect ­ region Asymptotic ­ number of sequences

Num ber of sequences to cover 100% antigenic diversity

10

8

6

4

2

0 0 1 2 3 4 5 Region of envelope reference sequence

Percentage of antigenic diversity covered by various percentages of unique sequences for HA proteins of influenza A virus

100

Percentage of antigenic diversity coverage (%)

80

60

H1 H2 H3 H5

40

20

0 0.00

20.00

40.00

60.00

80.00

100.00

Percentage of UNIQUE sequences (%)

Reduction in number of antigenically unique human HA sequences

1400

Number of sequences

1200 1000 800 600 400 200 0

All

Human

Human Unique Human Unique Antigen. H1 H2 H3 H5 H7 H9

V. Brusic 2007

Reduction in number of HA targets All Human Human unique 1102 40% Human antigenically unique 607 22%

2771 100%

1509 54%

Large reduction, but still too many proteins for a systematic study

V. Brusic 2007

The reduction of number of sequences (by observing antigenic redundancy) is significant, but still too high for vaccine studies Some regions of antigens contain high concentration of peptides that can bind to multiple HLA alleles that belong to the same supertype ­ T-cell epitope hot-spots

(Brusic et al., Immunol. Cell Biol. 2002, Srinivasan et al. Bioinformatics 2004)

These peptides are 15-50 amino acids long Ideal for further reduction of number of vaccine targets

V. Brusic 2007

MULTIPRED ­ prediction of T-cell epitope hot-spots HLA-A2, -A3, and -DR

(A)(C)

Analysis of prediction accuracy for five human proteins experimentally tested for HLA-DR hot-spots. Protein HGH EXP 34-48 97-132 181-195 Calcitonin Insulin IFN B 27-51 77-101 132-161 162-187 Epo 73-107 118-147 158-182 PRED 35-48 101-132 180-208 35-50 22-38 78-96 107-122 140-160 164-187 96-110 163-183 % MATCH 93 89 100 48 76 70 92 34 80 FP TP/FN TP FP TP TP TP/FN FN TP Thr 75 TP TP TP

7 TP, 2 TP (partial), 1 FP, 1 FN

V. Brusic 2007

VAGAT ­ Viral AntiGenome Analysis Tool

Protein H1

No. of antigenically unique sequences 180

Hot spots for HLA-DR 1 ­ 21 188 ­ 204 213 ­ 229 264 ­ 283 323 ­ 337 522 ­ 555 Total 1 ­ 22 157 ­ 172 184 ­ 200 518 ­ 553 Total 1 ­ 16 191 ­ 205 209 ­ 223 257 ­ 273 531 ­ 563 Total 1 ­ 19 159 ­ 177 186 ­ 202 321 ­ 341 527 ­ 559 Total

Hot spots Length (aa) 21 17 17 20 15 34 124 22 16 17 36 91 16 15 15 17 33 96 19 19 17 21 33 109

Binding Prediction 83.77 77.62 79.05 77.53 76.42 82.69 90.19 76.88 76.62 83.74 83.92 76.13 75.40 76.07 80.82

Hotspot Rank 1 4 3 5 6 2 1 3 4 2 1 3 5 4 2

H2

49

H3

355

H5

12

84.71 76.42 76.98 76.04 80.77

1 4 3 5 2

HLA Supertypes HLA-A2 HLA-A3 HLA-DR

Total no. of minimal peptides 62 175 262

HA FINAL No. of nonoverlapping minimal peptides* minimal peptides 47 47 146 144 190 188

The positions of T-cell epitope hot-spots are conserved even in proteins that show no sequence similarity Indicates structural properties that determine clusters of T-cell epitopes

Data Bioinformatics Hypothesis Experimental validation

V. Brusic 2007

We can significantly reduce the number of vaccine targets by focusing on T-cell epitope hot-spots Influenza A hemagglutinin T-cell epitope hot-spots differ between HLA supertypes and some do subtypes do not have hotspots (eg. H3 A2 and A3) T-cell epitope hot-spots from this study range in length from 15 to 93 AA, with average length of 21 AA (HLA-DR), 47 AA (A2) or 50 AA (A3) Class I HLA hot-spots are longer and have fewer predicted individual peptides

V. Brusic 2007

Analysis of PB2

Shannon Entropy Mutual information

H ( x) = -

eE

pe ( x) log2 (pe ( x))

H S ,a ( x ) = -

MI ( x) = H a ( x) + H S ( x) - H S ,a ( x)

H S ( x) = -

n1 n n n log 2 1 - 2 log 2 2 N N N N

p(S , a) log2 p(S , a)

S aA

PB1 binding NP binding RNA cap binding NLS A2A variants

DE

9

A

44

M

64

T TA

81 105

A

199

TI IV

271 292

R

368

L

475

DE AV VA E

567 588 613 627

A AS

K

661 674 702

H2H variants

NT

S

T MV VM

S

A

T

K

M

N

I

TI

K

T

T

R

Chen (2006) Naffah (2000)

Miotto et al. in preparation

SUMMARY Introduction Immunological databases Analysis of viral sequences Computational models Conclusions and future developments

V. Brusic 2007

Variety of classification methods Motifs Matrices Artificial Neural Networks Hidden Markov Models Bayesian Networks Support Vector Machines Molecular Modelling Combined Models

V. Brusic 2007

Models in immunology Molecular level

Antigen processing and presentation Molecular interactions

System level

Modelling immune responses

Motta S. and Brusic V. (2004). Mathematical Modelling of the Immune System. In G. Ciobanu, G. Rozenberg (eds.) Modelling in Molecular Biology, 193-218, Natural Computing Series, Springer, 2004.

Pathogen diversity

V. Brusic 2007

Model requirements

High accuracy

High specificity (cheap confirmation) High sensitivity (broad coverage)

Generalisation

Predict well previously unseen peptides Predict well across allelic variants

Improvement over time Robustness (resistance to errors and biases)

V. Brusic 2007

Example 1

1994 - Prediction of MHC class I binding peptides Molecule: HLA-A*0201 Subset: 9-mers Data: 186 binders, 1071 non-binders Model ­ Neural networks

V. Brusic 2007

V. Brusic, 2002

Example 1

Predicted vs. experimental binders

80 70 60 50 40 30 20 10 0 Predicted Binders Predicted Non-binders Exp. Binders Exp. Non-binders

SE = 74%, SP = 82%

V. Brusic 2007

Brusic et al., Complex Systems, 1994

Example 2

1998 - Prediction of MHC class II binding peptides Molecule: HLA-DRB1*0401 Core: 9-mer Data: 338 binders, 312 (578) non-binders Models ­ Matrices and Neural networks

V. Brusic 2007

Example 2

Experimental testing of protein thyrosine phosphatase (IA-2) in at-risk IDDM relatives Binding assays T-cell proliferation assays

Honeyman et al., Nat. Biotechnol. 1998 Brusic et al., Bioinformatics 1998

V. Brusic 2007

.

Example 2

HLA-DR4 T-cell epitopes from an IDDM antigen IA-2

1000

T-cell resp. < 1 SD T-cell resp. 1-2 SD

Binding Index ( 1/IC50)*100

T-cell resp. > 2 SD

100

10

1

-2

0

2

4

6

8

10

Binding Prediction

V. Brusic 2006

Ex. 2

Predicted vs. experimental binders

Pred. Binders

1.00 0.80 0.60 0.40 0.20 0.00 Exp. Binders Exp. Non-binders

Pred. Non-binders

V. Brusic 2006

Fraction of total

Example 2

Predicted and experimental binding as predictors of T-cell epitopes

T-cell epitopes 1.00 Fraction of total 0.80 0.60 0.40 0.20 0.00 Pred. binders

V. Brusic 2006

Missed T-cell epitopes

Exp. Binders

Example 3: cyclical refinement

Initial experiments refine

Optimise/ clean

Computer models

Further experiments

define

V. Brusic 2007

Example 3

Malaria ­ 0.5 M cases per annum

Search for vaccine targets in HLA-A11 population in Vosera - Papua New Guinea Six antigens from P. falciparum

LSA-1 SALSA CSP GLURP STARP TRAP ~1909 AA ~ 83 AA ~ 432 AA ~1262 AA ~ 604 AA ~ 559 AA

3127 peptides

V. Brusic 2007

Example 3

1)

Overlapping study

Twenty overlapping 9-mer peptides from the known immunogenic region of LSA-1

90 94 105

88 NVKNVSQTNFKSLLRNLGVSENIFLKEN 115 2) Initial ANN model: 98 binders and 145 non-binders

34 peptides selected and tested for HLA-A*1101 binding

3)

Refined ANN model: 123 (98+13+12) binders and

203 (145+41+17) non-binders twenty-nine (29) peptides were selected and tested

V. Brusic 2007

Correctly predicted binders

100 80

3/20

10/36

22/29

%

60 40 20 0

15 29 76

Overlapping peptides

V. Brusic 2006

ANN 1st round

ANN refined

Brusic et al. Journal of Molecular Graphics and Modelling, 2001

Various applications

Relationship between binding to TAP transporter and MHC Prediction of cancer-related T-cell epitopes Identification of autoimmunogenic T-cell epitopes Prediction of peptides that bind multiple MHC molecules Large-scale (genome-wide) screening of MHC binders Prediction of renal transplant outcomes Identification of viral vaccine targets

V. Brusic 2007

We have developed a number of integrated systems MULTIPRED HotSpot Hunter PREDBalb/c PREDNOD Molecular modelling for HLA-DQ and HLA-C VAGAT Emphasis on experimental validation

V. Brusic 2007

MULTIPRED ­ prediction of T-cell epitope hot-spots HLA-A2, -A3, and -DR

(A)(C)

Need for models/simulators

It is impossible to perform systematic studies given the combinatorial complexity of the immune system and its targets Computational analysis is used to select "the best" or "key" experiments Cyclical refinement accelerates discovery process Computational models can be as accurate as experimental procedures, however this requires a significant testing and validation effort

V. Brusic 2007

THE "TRIPLEX" VACCINE

De Giovanni et al., Cancer Res. 64: 4001, 2004

Allo-MHC (H-2q)

IL-12 genes

p185neu

IL-12

Slide courtesy of Santo Motta

THE SIMTRIPLEX SIMULATOR

100

Tumor-free (%)

80 60 40 20 0

Chronic Early Late Very late Untreated

Triplex vaccine in real mice

0

10

20

30

40

50

60

70

80 Weeks of age

Atypical hyper- CIS Tumor plasia

Multiple metastatic tumors

100

Tumor-free (%)

80 60 40 20 0 0 10 20 30 40 50 60 70

Atypical hyper- CIS Tumor plasia Multiple metastatic tumors

Chronic Early Late Very late Untreated

SimTriplex in virtual mice

80 Weeks of age

(Pappalardo et al., Bioinformatics, 2005) Slide courtesy of Santo Motta

USING SIMTRIPLEX TO FIND OPTIMAL/ MINIMAL VACCINATION SCHEDULES

1. Heuristic approach Based on the "Early" module, a posteriori driven by Cancer Cells number. Tumor-free mice at one year: 96%. Number of vaccinations reduced by 27% in comparison to "Chronic" protocol. Slide courtesy of Santo Motta

USING SIMTRIPLEX TO FIND OPTIMAL/ MINIMAL VACCINATION SCHEDULES

2. Genetic algorithm Driven by SimTriplex outcome (survival >400 days). Fitness function: - minimize number of vaccinations; - keep Cancer Cells kinetics similar to "Cronic" schedule

Slide courtesy of Santo Motta

USING SIMTRIPLEX TO FIND OPTIMAL/ MINIMAL VACCINATION SCHEDULES

Tumor-free mice

Early

Vaccinations 12 12 12 60 (100%) 44 (-27%) 35 (-42%)

0% 0% 0% 85%-100% 75%-96% 84%-91%

10 20 30 40 50 60 70

In vivo

Late Very late Chronic

In silico

Heuristic Genetic 0

Weeks of age

Slide courtesy of Santo Motta

SUMMARY Introduction Immunological databases Analysis of viral sequences Computational models Conclusions and future developments

V. Brusic 2007

Database

Experiment

V. Brusic 2007

DEALING WITH COMBINATORIAL COMPLEXITY OF THE IMMUNE SYSTEM Database In silico experiment Targeted experiment

V. Brusic 2007

Data warehousing provides a framework for this process, And computational models provide for rapid pre-screening

CRITICAL ISSUES Modelling capabilities are quite advanced but they need to be deployed correctly and take into account biological concepts For example, simple combining predictions of proteasome cleavage, TAP binding, and MHC class I binding to find "best candidates" is not very useful because there are alternative antigen processing pathways Diagnostics needs to be improved so that molecular subtypes in both clinical and pre-clinical studies are well-defined

V. Brusic 2007

basic immunology maths/stats

clinical immunology molecular biology

artificial intelligence

IMMUNOMICS

cell biology

databases

V. Brusic 2007

algorithms systems science

physics/chemistry

GROWTH AREAS Humoral immunity and deciphering antibody-antigen interactions Large-scale integrative projects, which are treated Favorably by granting bodies Human Physiome Project Human Immunome Project? Cancer Immunome Project?

V. Brusic 2007

R&D FUNDING

1994-1998 Walter and Eliza Hall Institute and Len Harrison, Melbourne, Australia 1999-2001 (US$0.4M) National Science and Technology Board of Singapore for the project Computational Immunology 2002-2005 (US$ 2.8M) Core funding Institute for Infocomm Research, Singapore 2003-2007 (US$7.5M) NIH ,1 U19 AI56541-01, a multi-project grant. Project 1, Computational identification of dengue virus T-cell epitopes 2004-2008 (US$7.2M) Contract HHSN266200400085C (NO1-AI40085) ­ Large Scale Antibody and T-Cell Epitope Discovery Program 2006-2008 ( 2M) IST-04-028069-STP. Commission of the European Communities, Information Society Technologies, Specific Targeted Research Project, ImmunoGrid ­ The European Virtual Human Immune System Project 2007-on Cancer Vaccine Center, Dana-Farber Cancer Institute

V. Brusic 2007

More than 100 people actively contributed to the work presented in this talk I am grateful to my mentors, collaborators, students and staff

V. Brusic 2007

Information

Microsoft PowerPoint - ImmunoiformatincsHarvard2007.ppt

92 pages

Report File (DMCA)

Our content is added by our users. We aim to remove reported files within 1 working day. Please use this link to notify us:

Report this file as copyright or inappropriate

469964