1. Rational vaccine design using reverse vaccinology, database data analysis, and literature mining
Yongqun “Oliver” He Unit for Laboratory Animal Medicine Department of Microbiology and Immunology Center for Computational Medicine and Bioinformatics
Comprehensive Cancer Center
University of Michigan Medical School Ann Arbor, MI 48109

2. Outline  Introduction
Vaxign: Rational vaccine design using reverse vaccinology
Rational vaccine design based on database data analysis
Rational vaccine design by literature mining
Summary and Discussion

3. “We must make this the decade of vaccines.”
Infectious diseases resulted in 9.2 million deaths in 2013 (about 17% of all deaths)
― Global Burden of Disease (GBD) Study 2013
“We must make this the decade of vaccines.”
-- Bill Gates (2010)
Bill Gates pledged $10 billion for vaccines for the “decade of the vaccine.”

4. How to rationally design vaccines? Can bioinformatics help?

5. Outline
Introduction
Vaxign: Rational vaccine design using reverse vaccinology
Rational vaccine design based on database data analysis
Rational vaccine design by literature mining
Summary and Discussion 

6. Reverse Vaccinology (RV)
Vaccine development starts with bioinformatics genome sequence analysis
Dr. Rino Rappuoli
Pioneer of
Reverse Vaccinology
Neisseria meningitidis serogroup B
 meningococcal B (MenB) disease
Pizza M, et al. Science.
;287(5459):
Milestone:
Novartis’s vaccine Bexsero
approved by Europe & USA
- Johri et al. Nat Rev Microbiol Dec; 4(12):932–42.
- Rappuoli R. Curr Opin Microbiol Oct;3(5):

7. Vaxign: Vaccine Design System
The 1st web-based reverse vaccinology system
Predicted criteria by Vaxign pipeline:
Antigen subcellular localization (original criterion)
Adhesin (critical for host invasion)
Transmembrane domains
Epitope binding to MHC class I and class II
Pathogen gene ortholog analysis
Sequence similarities to human, mouse and/or pig proteins
Freely available:
Reference: He Y, Xiang Z, Mobley HLT. Vaxign: the first web-based vaccine design program for reverse vaccinology and an application for vaccine development. Journal of Biomedicine and Biotechnology. Volume 2010 (2010), Article ID , 15 pages. [PMID: ]

8. Two Forms of Vaxign Usage
Pre-computed data query
Dynamic analysis
Runtime execution
Can analyze up to 500 proteins at one time
Pre-computed results
> 200 genomes in database
Easy to query

9. Brucella Vaccine Design using Vaxign
From 2 months of work with computational experience, now we only need less than one minute. Gr

10. Brucellosis is a zoonotic disease caused by bacterium Brucella
Vaxign case study:
Brucellosis is a zoonotic disease caused by bacterium Brucella
Brucellosis in livestock
Human brucellosis
Global Livestock Production and Health Atlas, Animal Production and Health Division, 2006, FAO
Ref: Pappas G, et al. Lancet Infect Dis Feb;6(2):91-9.
>500,000 new human cases worldwide annually

11. Brucella Genomes for Vaccine Design
Species
Host
B. abortus
cattle, human
B. canis
dogs, foxes, coyotes, human
B. melitensis
sheep, goats, human
B. neotomae
desert wood rats
B. ovis
rams
B. suis
pigs, human
4 new Brucella spp. :
B. cetacea
cetacean
B. pinnipedia
seal
B. microti
voles
B. inopinata
unknown
10 Brucella spp.
4 pathogenic to humans
10 genomes in NCBI RefSeq database
> 20 new genomes were sequenced
Question: Can we use Vaxign to predict Brucella vaccine candidates?

12. Vaxign Prediction of Brucella vaccine targets
*, represents five genomes from virulent B. abortus strain 9-941, B. melitensis strains 16M and ATCC 23457, and B. suis strains 1330 and ATCC
**, represents the genome from B. ovis strain ATCC 25840;
The indicated two proteins are Omp2b (YP_ ) and Omp31-1 (YP_ ).
Four virulent Brucella strains in 3 Brucella spp. are studied. We look for the conserved genes among all these strains.
Ref. He Y, Xiang Z. Bioinformatics analysis of Brucella vaccines and vaccine targets using VIOLIN. Immunome Research Sep 27;6 Suppl 1:S5.

13. Experimental Verification of Vaxign-predicted Brucella vaccine candidates
Experiments by Dr. Thomas Ficht’s lab in Texas A&M:
5 Vaxign-predicted proteins expressed & tested
2 proteins with the highest predicted adhesin scores conferred an invasive phenotype to the non-invasive BL-21 E. coli strain in alveolar epithelial cells
All proteins elicited IgG antibody production in Brucella-exposed goats, mouse and humans.
3 proteins elicited protection in mice against challenges
This is a great demonstration of Vaxign usefulness
Reference: Gomez et al. Immunogenicity and protective efficacy of Brucella outer membrane proteins identified via a reverse vaccinology approach Annual Brucellosis Conference, Chicago, IL, USA.

14. Vaxign-Vaxitop MHC Class I and II Epitope Predictions
Internally developed project
Based on position specific scoring matrices (PSSM)
Unique: Calculate statistical P-value for each prediction
Also a winner of the 2nd Machine Learning Competition in Immunology 2012:
Task: “Predict peptides naturally processed by MHC Class I pathway ("eluted peptides") for each target MHC molecule”
Winning category: Group 1: B*5701

15. Other Vaxign Applications
Uropathogenic E. coli (UPEC) vaccine prediction
Collaboration with Dr. Harry Mobley here in UM
African Swine Fever Virus vaccine development
Collaboration with Mangkey Bounpheng et al.
All predicted & expressed proteins were found immunogenic
More experiments under way
Human herpes simplex virus vaccine design
Francisella tularensis vaccine design
Vaxign can also be used for virulence factor prediction (e.g., Campylobacter fetus, Corynebacterium diphtheriae)
~50 citations of Vaxign in Google Scholar
References:
He Y, Xiang Z, Mobley HLT. Vaxign: the first web-based vaccine design program for reverse vaccinology and an application for vaccine development. J Biomed Biotechnol. 2010;2010:
Xiang Z, He Y. Genome-wide prediction of vaccine targets for human herpes simplex viruses using Vaxign reverse vaccinology. BMC Bioinformatics. In press.

16. What’s More about Vaccine Design
Vaxign is to predict protective antigens,
which can be improved using other data, including database and literature data
In addition
Need to design more than protective antigens, e.g., DNA vaccine plasmid, vaccine vector, adjuvant, gene for deletion, …

17. Outline
Introduction
Vaxign: Rational vaccine design using reverse vaccinology
Rational vaccine design based on database data analysis
Rational vaccine design by literature mining
Summary and Discussion 

18. VIOLIN: Vaccine Investigation and Online Information Network
Comprehensive vaccine R & D database & data analysis system
>3400 vaccines (licensed, clinical trial, or lab-verified)
200 pathogens or non-inf. diseases (e.g., cancer)
~1200 pathogen genes:
Protegen: Protective antigens
Virmugens: Genes mutated to make live attenuated vaccines
>1,000 host genes
Databases of different vaccine components
Vaxjo: Vaccine adjuvants
Vaxvec: Vaccine vectors
DNAVaxDB: DNA vaccines and DNA vaccine plasmids
> 2,000 papers manually annotated

19. Protegen: Protective Antigen Database
Protective antigens (Ags): pathogen Ags tested in experimentally verified vaccines
857 protective Ags collected in Protegen
Enriched patterns among protective Ags found
Secreted and outer membrane proteins favorable, depending on pathogens
Many conserved domains were identified, for example,
Protegen data have been used by many as “Gold Standard” for testing
Reference: Yang B, Sayers S, Xiang Z, He Y. Protegen: a web-based protective antigen database and analysis system. Nucleic Acids Research :D

20. Can we predict protective antigens using Protegen database data?
Protegen data can be used to identify enriched patterns, which can guide rational vaccine design
E.g., from analysis of 270 bacterial protective antigens
Extracellular or cell wall proteins: 64% for Gram+
Extracellular or outer membrane: 48% for Gram-
Adhesins or adhesin-like: 54% G+; 40% G-
Enriched conserved domains, e.g., TonB
A Support Vector Machine (SVM) classification: 92% true positive rate of sequence-based protection
Reference: He Y, Xiang Z. Bioinformatics analysis of bacterial protective antigens in manually curated Protegen database. Procedia in Vaccinology Vol 6. Pages 3-9.

21. “Virmugen”: virulence factors whose mutants are live attenuated vaccines
VirmugenDB: database that collects and analyzes virmugens
URL:
Statistics: 225 virmugens from 57 pathogens; 196 vaccines
16 “aro” genes: for aromatic amino acid synthesis
Enriched functions: Nutrient metabolism & Cell membrane formation
Support rational live attenuated vaccine design
Reference: Racz R, Chung M, Xiang Z, and He Y. Systematic annotation and analysis of “virmugens” - virulence factors whose mutants can be used as live attenuated vaccines. Vaccine Jan 21; 31(5):

22. Vaxjo: Vaccine adjuvants
93 vaccine adjuvants
Used in 379 vaccines
Question: Does the information help adjuvant selection in vaccine design?

23. Vaxvec: Recombinant Vaccine Vectors and recombinant vector vaccines
59 recombinant vectors
187 recombinant vector vaccines
Bacterial antigens in Vaxvec
Adhesins: 19%
Compare DNAVaxDB: 30-39%
Microbes used as recombinant vaccine vectors:
Question: Does the information help recombinant vector vaccine design?

24. DNAVaxDB: DNA vaccines and plasmids
Bacterial DNA vaccine antigens:
Extracellular: G+: 64%; G-: 16%
Adhesins: G+: 39%; G-: 30%
420 verified DNA vaccines
181 DNA vaccine plasmids
Plasmids with >=5 associated vaccines:
Question: Does the information help DNA vaccine design?

25. Ex: Develop DNA Vaccines for M. bovis
Mycobacterium bovis:
Aerobic bacterium that can cause tuberculosis, primarily in cattle, also in humans
part of M. tuberculosis complex
Not well studies: in VIOLIN, only BCG and pCI-Ag85B
May use M. tuberculosis or other vaccines information
Plasmid selection:
Commonly used plasmids in M. tuberculosis: pJW4303, pcDNA3.1.
Some plasmids are used more in Gram +/- bacteria and viruses
Protective antigens:
Used in M. tuberculosis DNA vaccines: Ag85A, KatG, MPT64  find orthologs in M. bovis.
Prediction using Vaxign

26. Outline
Introduction
Vaxign: Rational vaccine design using reverse vaccinology
Rational vaccine design based on database data analysis
Rational vaccine design by literature mining
Summary and Discussion 

27. Can literature mining support rational vaccine design?
Publications
in PubMed:
Vaccine publications
have increased exponentially
Reference: He Y, Rappuoli R, De Groot AS, Chen RT.
Emerging vaccine informatics. J Biomed Biotechnol. 2010;2010:

28. Ex. Search genes associated with “live attenuated Brucella vaccine” in PubMed
Method:
Default
MeSH-based PubMed
search

29. Vaccine Ontology (VO) defines more live attenuated vaccines
VO: Open source ontology for vaccines
Inferred hierarchy
Asserted hierarchy

30. Ex. Search genes associated with “live attenuated Brucella vaccine” in PubMed
Method:
VO-based SciMiner
PubMed
search

31. Ex. Search for genes associated with Brucella vaccine in PubMed
Method: VO-SciMiner
VO includes and classifies 67 Brucella vaccine
~15,000 PubMed abstracts searched
140 Brucella vaccine related genes identified
Functions of these genes studied:
protective antigens
virulence factors
other genes closely related to Brucella.
Some genes useful for vaccine development
Reference: Hur J, Xiang Z, Feldman EL, He Y. Ontology-based Brucella vaccine literature indexing and systematic analysis of gene-vaccine association network. BMC Immunology. 12(1): Aug 26. PMID:

32. Summary & Discussion
Vaxign: Vaccine design by predicting protective antigens through bioinformatics analysis of genome sequences (i.e., reverse vaccinology)
VIOLIN vaccine component databases can support design of vaccine components
Literature mining also supports rational vaccine design
Discussion: What’s more about vaccine design?
Using OMICS data?
More or better software?

33. Acknowledgements He Lab at UM: U. of Michigan (UM): VO Collaborators:
Zuoshuang Xiang
Asiyah Yu Lin
Andrew Hodges
Rebecca L. Racz
Samantha Sayers
Monica Chung
Omar Tibi
Mukti A Patel
Shelley Zhang
Desikan Jagannathan
Joseph Alan Ostrow
Erica Laura Marcos
G. Bill Jourdian
U. of Michigan (UM):
Junguk Hur
Arzucan Ozgur
Dragomir R. Radev
Brian Athey
Harry Mobley
VO Collaborators:
Barry Smith (U Baffalo)
Lindsay Cowell (UT SW)
Bjoern Peters (LIAI)
Alan Rutternberg (Baffalo)
Melanie Courtot (Canada)
Richard Scheuermann
(JCVI)
Funding:
NIH-NIAID Grant 1R01AI081062
NIH/NIAID 1 R21 AI057875
UM CCMB pilot grant
NIH 1 U54 DA021519