YouTube Facebook LinkedIn Google+ Twitter Xinginstagram rss  

Genomics, bioinformatics, and novel laboratory techniques are converging to boost vaccine research against a new wave of emerging diseases, natural and man-made. Now, will in silico modeling ramp up sufficiently to further speed vaccine discovery?

By Malorye Branca

June 15, 2003 | In a curious irony, the springtime celebrations for genetics and the human genome were gate-crashed by some new and familiar foes from the world of infectious diseases, casting fresh urgency on the hitherto unglamorous world of vaccine design.


*In April, barely two months after SARS (severe acute respiratory syndrome) emerged from China to become a worldwide health threat, the causative corona-virus had been unmasked and its entire genome sequence posted on the Web. Just a month later, in silico modelers had generated a list of parts from the SARS genome that are candidate vaccine ingredients.

*Also in April, scientists at St. Jude Children's Research Hospital in Memphis, Tenn., announced that in a period of just four weeks, they had created a candidate vaccine against a deadly strain of avian influenza virus that struck Hong Kong twice in the past few years.

* The World Health Organization announced that it would soon have a test kit available for another new form of avian flu, which broke out in the Netherlands in February; a vaccine for that strain should quickly follow.

Bullseye: Decoded microbes such as anthrax are spurring the design of vaccines.

Source: The Institute for Genomic Research
The gratifying progress against these new microbial menaces hinges on a widely heralded combination of data, computer power, molecular biology, and analytical software. Here, genomes are plentiful and growing ever more so. The IT power exists not only to rapidly analyze and compare the pint-sized sequences, but also to track them down and model their spread when they become public health threats. Steady progress on the laboratory side has made it easier to grow a wider range of microorganisms and to manipulate their genomes, cutting and pasting genes with razor-sharp precision.

Now, even the Holy Grail of drug design — in silico modeling — is starting to pay off for vaccine research. Here is where the greatest work is still required, but many believe it also holds the key to providing vaccine development its next great boost.

An Explosion in Interest 
All this technical progress is making the work of vaccine researchers dramatically simpler, and helping draw more interest in this once moribund field.

"Vaccines have had a bad rap," says Susan Burgess, president and CEO of startup Plexus Vaccine, "because they have not been very profitable, and there is a lot of risk in developing childhood preventive vaccines."

The SARS Effect 
With 3,000 children in Africa dying of malaria each day, and tuberculosis and AIDS inflicting similar damage, no wonder some people are asking: "What is all the fuss about SARS?"

Read More 
Powerful forces are rapidly changing that perception. "All this bioterror stuff is generating huge interest in the 'horror' bacteria," says Darren Flower, of the Edward Jenner Institute for Vaccine Research, in Berkshire, England. "Ten years ago there may have been 10 to 15 companies active in vaccine research; now there are hundreds."

The emergence of deadly new microbes, such as avian influenza, SARS, and West Nile virus, in rich nations as well as poor, is attracting commercial attention to the field. Not surprisingly, as demand rises, companies should be able to charge more for vaccines, according to Jennifer Cassels, an analyst at Frost & Sullivan. The consultancy predicts the global market for prophylactic vaccines will double over the next 10 years, reaching more than $10 billion.

Already, Merck and GlaxoSmithKline have announced SARS vaccine initiatives. A growing pack of smaller companies are also working on this or other new vaccines.

Sequencing Secrets 
At the root of this remarkable confluence of global trends and high technologies lies the genomic revolution. Vaccine developers were essentially working blind before microbial genomes started streaming out of sequencing centers. The first viral sequence was obtained by Fred Sanger in the United Kingdom in 1977. Craig Venter's team at The Institute for Genomic Research (TIGR) sequenced the first bacterial genome — Haemophilus influenzae — in 1995, sparking an explosion of bacterial sequencing. Today, more than 100 bacterial genomes have been sequenced, with more than 300 pending (see

Before this revolution, most vaccines consisted of a crippled version of the actual pathogen. However, these "attenuated" vaccines could, very occasionally, result in a full-blown infection, and were difficult to manufacture.

Suppressed enthusiasm: Christian Loucq of Acambis says it's too soon to understand the role of molecular modeling in vaccine design.
"We were taking viruses or bacteria, growing them, randomly inactivating them, and attenuating them by a couple of passages," says Christian Loucq, vice president of sales and marketing at Acambis, a vaccine developer, "but without ever knowing exactly what we were doing."

Now, each sequenced microbial genome spills out valuable secrets. "Once you have access to all an organism's genes, you are no longer missing any of the pieces of the puzzle," says Hervé Tettelin, an associate investigator at TIGR.

"Today, when we attenuate bacteria, we know exactly what kind of modification we are doing," Loucq says. "Now, not a single scientist would come up with a vaccine without knowing what they had changed and why." Besides being able to tinker more precisely with whole pathogens, researchers can select and string together bits and pieces of a microbe's genome — just enough to make the immune system recognize, and shut down, any possible infection.

It's called "rational" vaccine design.

The torrid pace of microbial genome sequencing is remarkable. In the past month, teams from TIGR and Integrated Genomics simultaneously published in Nature the completed sequences for a strain of anthrax — Bacillus anthracis — and one of anthrax's less-virulent cousins, Bacillus cereus, respectively. (Read, T.D. et al. Nature 423, 81-86; 2003; Ivanova, N. et al. Nature 423, 87-91; 2003.)

The anthrax strain sequenced at TIGR was "virtually identical" to the one used in the 2001 terror mailings, and is the first anthrax strain to have been completely sequenced. Both groups' analyses revealed new genes that distinguish the lethal bacterium from its cousins.

The TIGR scientists also used a DNA microarray containing anthrax gene sequences to compare the bacterium to two of its supposed relatives, including B. cereus. The study suggests that it's not merely different genes that distinguish anthrax, but how some of those genes are regulated.

Circling in: It's easier to select "pieces" of a patho-gen needed to make a vaccine when you can compare strains to find commonalities and differences.
With all these tools for analyzing and manipulating genomes, finding the best ingredients for a vaccine has become almost an industrialized process. These advances are also helping with bugs that have defied vaccine developers for decades.

"For 30 to 40 years we were completely unable to find any candidate vaccine for Group B meningitis," Tettelin says. "Nothing at all came up that was worth pursuing." No single candidate vaccine was protective against both old and new outbreaks. Using genomics, in about 18 months Tettelin's group found several proteins that might do the job.

The group started by using bioinformatics to predict, based on sequence analysis, which of the bacterium's proteins rode its surface: That's the point where the human immune system "interacts" with the microbe, Tettelin says. A search through the pathogen's approximately 2,000 genes turned up 600 possible surface proteins.

Laboratory techniques, including microarrays, confirmed which among those were indeed expressed on the bacterium's outer coat and were also recognized by the immune system, whittling the number of candidate proteins down to 85 (Grifantini, R. et al. Nature Biotechnology 20, 914-921; 2002).

The next step was to check whether these genes were found in enough Group B strains to make a broadly acting vaccine. Knowing where the genes should lie, the scientists sequenced those particular stretches of DNA in 25 different strains, and used sequence-comparison software to verify whether the genes were actually present.

Just seven proteins were left standing after this final filtering process, but that's enough for a vaccine program. Chiron Corp., TIGR's collaborator on the project, now has a vaccine based on this work in Phase I clinical trials.

Rational combination: Plexus' Virtual Epitope algorithm predicts suitable epitopes for vaccine design, which are then combined with a carrier protein to create a virus-like particle (far right) that primes the immune system without giving rise to disease.

Predicting an Immune Response 
The next task is to reduce the amount of lab work needed even further, by predicting how the immune system will respond to, and physically interact with, a specific protein. It is an infinitely harder problem than predicting the proteins from a microbial gene sequence, but many academic groups and several companies are rising to the challenge.

Algorithms are helpful right from the point where the immune system first connects with the foreign invader. One way the immune system finds out about invaders is by continually chopping up proteins into tiny pieces called peptides, and examining these to see if they are native or foreign. Epitopes are peptides that are recognized by the immune system and generate a response.

Getting that particular switch to flip is the major challenge in vaccine design.

The immune system is quite selective about which kinds of peptides it will recognize: It is estimated that only about 1 in 100 fits the bill. Computational modelers are trying to learn the rules underlying this selection process and develop algorithms that will help them pick better proteins for vaccines.

"If you know how the immune system handles certain kinds of proteins, you should be able to apply that knowledge to the growing genomic databases, and to answer the question, 'How would the system react to this particular protein?'" says Søren Buus, a professor at the Institute for Medical Microbiology and Immunology at the University of Copenhagen.

Buus built such a program with neural network specialist Søren Brunak, director of the Center for Biological Sequence Analysis, at the Technical University of Denmark. Buus provided the binding strength and other data, and Brunak's team the algorithms. Their approach involves alternating rounds of training for "committees" of neural networks on the data with actual experimentation. "When the committees don't agree, that means the model for a particular peptide isn't very good. Then we take it back to Søren, and he does more experiments," Brunak says.

Their model predicts how the T-cell-mediated immune process will break proteins apart, and which of the resulting peptides will be recognized by the system. A recent study describes the model in detail, and how they used it to predict T-cell epitopes for the hepatitis C virus (Nielsen, M. et al. Protein Science 12, 1007-1017; 2003). The program is unique, Buus says, because it is "the first that predicts binding strength in quantitative terms." Other approaches use scores that give the relative binding strength.

Plexus Vaccines acquired the patent on the program last year. The company also has 3-D structure prediction and modeling tools from Mol-Soft, which they use to predict epitopes that can be recognized by other parts of the immune system, controlled by B cells. T cells recognize only small linear peptide fragments. Some B-cell epitopes are linear, but others are folded into more complex shapes.

Plexus signed its first deal with a large pharmaceutical firm, GlaxoSmithKline, last year. The two companies will "explore" various computational approaches to rational vaccine design together. On its own, Plexus has been developing therapeutic vaccines for chronic infections, such as drug-resistant tuberculosis and hepatitis C. Unlike traditional immunizations, therapeutic vaccines aim to treat disease, not prevent it. Earlier this year, Siga Technologies announced its plans to acquire Plexus to enhance its own vaccine platform. In what Burgess describes as a "fabulous" deal, Siga plans to use the new technology to expand its aggressive program of biodefense vaccines and therapeutics. The company specializes in infectious diseases and is now focusing on smallpox, anthrax, plague, and other potential bioterror bugs.

The new Siga/Plexus entity aims to quickly show the power of its approach. Buus' group downloaded the sequences of the SARS virus' proteins as soon as they were available and applied their algorithm to identify candidate epitopes for a SARS vaccine. They already have some of the epitopes predicted by the model in hand and are testing them. "Thanks to the deal with Siga, it's possible we could have a candidate vaccine in a year," Buus says.

In a bind: A T-cell epitope docked on an MHC Class I molecule. AlgoNomics has modeled about a dozen MHCs. Its software can analyze interactions of epitopes and receptors.
Belgium-based AlgoNomics is generating detailed models of MHC (major histocompatibility complex) receptors — intermediaries that lock onto epitopes and "present" them to T cells. A wide variety of these molecules exist, and because they are genetically determined, some are more common in the population than others. A good vaccine should have molecules that will stimulate the most common MHCs, so it can work in many different people.

AlgoNomics has built models of about a dozen of the best-known MHCs. The company's software can fit up to thousands of epitopes into multiple receptor models, and analyze their interactions. That is an intensive set of calculations, and the company uses a Fujitsu Siemens Computers' hpcLine cluster of 98 processors with 74 GB of RAM for the task. According to Chief Operating Officer Philippe Stas, AlgoNomics can screen multiple peptides against multiple receptors in just a couple of hours.

As soon as the SARS genome was posted, AlgoNomics' scientists also started running it through their models. By early May, they had generated a long list of SARS epitopes. Some of these, which bind the MHC-I molecule HLA-A*0201, are included in a list posted at and are available free for academic use.

CEO Ignace Lasters says that while AlgoNomics has the laboratory and software tools to find good epitopes, "We need partners who can actually make vaccines to work with us."

Epimmune is old enough to be considered a granddaddy in the computer modeling vaccine business. Based on research by Robert Chesnut and Alessandro Sette, Epimmune's computational tool kit includes algorithms that can pull out epitopes with specific configurations, modeling software to predict the strength with which an epitope will bind immune molecules, and another modeling program that can retrieve and analyze related epitopes.

The company is working on preventive vaccines for hepatitis C, HIV, and malaria, and is currently running a Phase I/II trial in both the United States and Africa for a therapeutic HIV vaccine. In collaboration with biotech Genencor International, Epimmune is developing additional therapeutic vaccines for hepatitis B, hepatitis C, and human papillomavirus.

Success in vaccine development "comes down to customization," says Mark Newman, Epimmune's vice president of infectious diseases. "There is a lot of variation [among strains], and with HIV, for example, you want a vaccine that you can use everywhere, not just in Africa." Hence, it is better to string together multiple, potent epitopes than to rely on just one.

In early May, Epimmune and Merck & Co. signed a deal reflecting the growing interest in vaccine modeling. The agreement calls for Merck to evaluate select Epimmune epitopes for possible vaccine development.

If they work properly, computational methods could take even more sweat out of the vaccine development process. But some experts have doubts about how far the field has advanced. "We are barely at the beginning with molecular modeling," Loucq cautions. Flower also notes that there is "much skepticism" about computational modeling among immunologists.

The weak spot in computational modeling is the paucity and unevenness of the data. "There are 500 or more MHC sequences alone, and only a small proportion have been analyzed," Flower says. The number of possible immunogenic peptides is even larger — more than 500 billion. Data has been captured for only a fraction of all the possible interactions between these two sets of molecules.

Studying all those interactions in the lab, however, would be a staggering task. "When you don't have data, often the best thing you can do is apply modeling," Flower says. Modeling specialists concede there are drawbacks to the approach, but they maintain that if modeling and experimentation are done in a complementary and iterative fashion, better answers can be gotten more quickly.

"You can't ever expect you will push a few buttons and you'll have a vaccine at the end of the day," says Juergen Hammer, global head of bioinformatics at Roche Pharmaceuticals in Nutley, N.J. An epitope prediction program he developed about 10 years ago, called Tepitope, is available free at

"With the genomic information available nowadays, you can do things you couldn't imagine before, like systematically going through all of an organism's proteome to predict all the potential T-cell epitopes," he says.

The models will improve as the amount of data on the interactions of immune molecules and epitopes grows. Flower's group, for example, has just released the JenPep database, which is freely accessible at This database contains a wealth of quantitative, experimentally derived data related to how strongly certain peptides bind to specific immune molecules.

Vaccines on Demand? 
In the face of the mounting spread of SARS or worries about the threat of a massive, airborne smallpox terror attack, there is intense interest as to just how quickly a vaccine can be delivered. Clearly, it will take more than some fancy new algorithms to drastically accelerate that process.

For one thing, many challenges remain on the biological front. Culturing the SARS virus, for example, proved to be a bigger hurdle than actually deciphering the minuscule 29,700-base genome. The ease with which a pathogen's genome can be manipulated also varies. The group at St. Jude, which developed the candidate avian flu vaccine, used a new twist on an established approach. Their technique delivered a potential vaccine quickly, but, says team leader Richard Webby, "The technology you have to use is specific for each family of virus. We could not use this with another type of virus, like SARS."

Instant vaccine: Researchers at St. Jude Children's Research Hospital used a premium mix of old and new viral influenza genes to create a possible avian flu vaccine in just four weeks.

Source: Richard Webby
HIV provides the starkest demonstration that microbes have many tricks up their sleeves too. It's been almost two decades since U.S. health official Margaret Heckler's notorious 1984 pronouncement that an AIDS vaccine would be available within two years. Despite tremendous progress, there is no certain sign that an effective vaccine is on the way.

Once discovered, vaccines must win approval, and that is by no means a formality. "It's been taking longer and longer for vaccines to get approved," AlgoNomics' Stas says, partly because vaccines are so effective that expectations have risen sharply, perhaps beyond the bounds of reason.

"Twenty years ago, no one would have been concerned about the rate of side effects from the [oral] polio vaccine," Loucq says. About 1 in 2.4 million people contracts the disease from the oral vaccine. In the face of tens of thousands of polio cases, the vaccine looks like a godsend despite the possible side effects. To anyone in a country where polio has been completely vanquished for more than a decade, however, the oral vaccine looks like a risky proposition. Luckily, a safer but weaker vaccine is now available, and it is used in regions where polio has been eradicated. In countries still fighting the disease, such as India, where polio has resurged causing 1,500 new cases last year, the oral vaccine is used.

The vaccine-versus-disease risk ratio would be completely reversed if a massive epidemic or terror attack occurred. Even the FDA, known for its thoroughness, is ready to bend the rules in case of truly catastrophic circumstances. A new FDA rule would allow drugs or vaccines against "lethal or permanently disabling toxic substances" to go directly from animal models to the public, "when the traditional efficacy studies in humans are not feasible and cannot be ethically conducted."

As a result of this curious mix of events, vaccine development is being transformed from a slow, imprecise science and commercial backwater into a truly global priority. The result should be not only the rapid-fire development of a handful of vaccines that ease the public's latest fantastical fears, but also genuine progress toward defeating many of the world's plagues, including the most widespread and endemic diseases — malaria, tuberculosis, and HIV.* 


For reprints and/or copyright permission, please contact Angela Parsons, 781.972.5467.