When scientists first published the initial results of the human genome in 2001, we seemed to be on the precipice of a revolution in medicine. Researchers could finally discover specific genetic mutations that lead to diseases. Pharmaceutical companies could devise scores of new drugs to target those mutations. Patients could be treated based on their individual genetic patterns.
So far, though, some scientists say the results have been disappointing. True, sequencing has become increasingly inexpensive, and this has led to a trove of data. Some of that data has provided important information to help prevent diseases; for instance, women who have the BRCA mutation and an increased likelihood of breast or ovarian cancer may be either carefully monitored or undergo a prophylactic mastectomy. Scientists have been able to use genetic sequencing to identify and group previously unknown diseases, which can lead to improvements in treatment.
But the flood of new drugs based on the genome hasn’t arrived. “It’s been an unmitigated failure from my perspective,” says Joseph Loscalzo, head of the department of medicine at Brigham and Women’s Hospital and Harvard Medical School, referring to so-called genome-wide analysis studies that mine genetic data in the search for new targets for drugs.
Loscalzo admits that the scientists involved may disagree with his forceful statement, but he points out that very few diseases have been linked to only one or a small number of genes. Even ones that are classically tied to a person’s genetics, such as sickle cell anemia, can cause vastly different symptoms, due, says Loscalzo, to the “genetic context” in which the mutated gene operates. For more complex and prevalent diseases, like cardiovascular disease—Loscalzo’s specialty—few patients have the same genetic variation or even the same symptoms.
The problem, Loscalzo says, is that the approach is too simple. Our genomes are just one part of a vast network of interactions in our cells and in our bodies. The way to treat medicine in the future will rely on a new approach, he says, one that scientist are now beginning to flesh out: network medicine.
The Creation of the Network
László Barabási was visiting a Chicago playground in 2000 with his young son when he met fellow Hungarian Zoltan Oltvai. Barabási, then a professor of physics at Notre Dame, had a professional interest in network theory; he would soon co-author an influential paper and publish a best-selling book called Linked: The New Science of Networks . Olvai was a Northwestern University cancer researcher.
Though he’d long been fascinated by biology, Barabási didn’t have the knowledge to delve deeply into the field. But on that Chicago playground, he and Oltvai began chatting. Oltvai suggested the two join forces.
“A mechanic will never be able to fix the car without knowing how the parts interact with the other components.”
Barabási was puzzled. “I said, ‘Well, where are the networks in biology?’ ” Barabási recalls himself thinking. But soon the pair published two important papers, one detailing the metabolic network in a cell, and another about the interactions among proteins.
Barabási began to focus deeply on biological networks. He met Marc Vidal, director of the Center for Cancer Systems Biology at the Dana-Farber Cancer Institute, at a network science meeting, and the two hit it off. Barabási went on to spend a sabbatical year at Vidal’s lab in 2005. During that time, he connected with Loscalzo, with whom he now collaborates. He then started his own lab at Northeastern University’s Center for Complex Network Research.
The genome hasn’t been a medical panacea, Barabási says, because it’s little more than a parts list, which can only get you so far. “A mechanic will never be able to fix the car without knowing how the parts interact with the other components,” he says.
That’s the premise of network medicine. All the parts of a cell interact in certain way. Sections of DNA code for proteins, which can interact with other proteins, transcription factors can turn on stretches of DNA, and so on. It’s these connections and pathways—what some scientists are calling the “interactome”—that could help us understand how genes lead to disease. There are potentially millions, even billions, of such connections. But thanks to clever experiments and vast computing power, a map is starting to emerge.
Mapping the Links
Vidal has been focused on detailing the interactions among proteins for more than 20 years. Back in the 1990s, he was trying to understand how the genome as a whole codes for physical attributes such as eye color or asthma. He developed the idea that a crucial missing link is the network of connections between proteins in the cell, which interact to help translate the genome into physical attributes.
At the time, he gave a presentation about this hypothesis. He imagined a map of such a wiring network, but, at the time, said it couldn’t be done. A colleague insisted that Vidal’s line of reasoning was too interesting to give up. “That night, I went to bed a little frustrated, thinking, ‘It’s a cool idea, but it’s impossible,’ ” Vidal says. “And the next day, I had an idea about how to solve it.”
Vidal was an expert in yeast genetics, as yeast cells provide a simplified model that helps scientists tease out challenging questions about human cells. He employed a method to test many protein interactions simultaneously and relatively quickly to determine how they interlock. It works something like this: He begins by inserting yeast cells with sections of human DNA. Those sections then build the associated human proteins, as they would in a human cell. If the yeast cell the proteins created by those two of those DNA sections link up, then the proteins also interlock in our bodies.
In 2005, Vidal published on the order of 1% of all possible human protein-protein interactions. This year, in a new paper currently under review, he and his team will publish approximately 20% of the expected full network. He’s since automated the system to speed the process and further improve the quality of the information. He hopes to produce a reference map of nearly the entire protein-protein network by 2020.
Barabási, Vidal, and Loscalzo have been publishing papers together on such networks and nodes—the genes and pathways where diseases tend to cluster. They call such an approach the “diseasome.”
Barabási likens the diseasome to a map of Manhattan: there are certain clusters where various activities take place—theater along Broadway, finance on Wall Street, advertising agencies on Fifth Avenue. The same is true for the patterns in a cell, though it doesn’t necessarily happen in physical space, the way neighborhoods cluster in a city. Rather, they play out in the chemical networks within a cell—specifically, which proteins, genes, and other chemicals are connected to one another. Using powerful computers to map these networks, researchers have been able to find regions where the connections that make up a particular disease “clump” together. They call these the “disease modules” in the network.
Scientists have been able to find some of these modules. In a presentation at Dana-Farber Cancer Institute, Barabási described how the team of researchers had analyzed 300 diseases. They tried to determine whether the genes that are known to have a connection to a particular disease link up to each other through gene and protein interactions.
Vidal hopes to produce a reference map of nearly the entire protein-protein network by 2020.
They found that 20% of disease-related genes form a connected network in a disease module. The other 80% are in the vicinity; they connect to the diseased genes through one non-diseased gene. (This may sound familiar to fans of the pop-culture game “Six-Degrees of Kevin Bacon;” the 80% of genes in the vicinity are two degrees separated from the disease, just like Johnny Cash is two degrees separated from Kevin Bacon—Cash was in All My Friends Are Cowboys with Willford Brimley, who appeared with Kevin Bacon in End of the Line .) That means that if, hypothetically, there are 100 genes known to have a connection to asthma, about 20 of them link to each other. For the other 80%, Barabási believes that scientists probably haven’t yet discovered the missing connections that would link them up to the disease module.
One surprising result that has come out of the networks thus far is an apparent relationship between asthma and celiac disease. Barabási and his colleagues demonstrated that there are similar pathways and significant overlaps in the wiring between the two diseases. Barabási says there’s also some clinical evidence that if you have one, you’re more likely to have the other.
Aviv Regev, director of the Cell Circuits Program at the Broad Institute, has taken a different approach to understanding networks and diseases. Instead of building up known connections of proteins and genes that might occur in any cell, her team conducts a deep analysis of one type of cell. The group attempts to model just that one cell type’s wiring system, encompassing all the different chemicals and systems that link up, along with the non-coding elements of DNA. First, they focused on one variety of cell in our immune system that, when it malfunctions, is implicated in conditions including autoimmune diseases such as multiple sclerosis and inflammatory diseases such as Crohns and ulcerative colitis.
It took years to understand all the connections in just that one cell. “But we developed the models from scratch,” Regev says. “Now the system is faster.” She adds, “We learned about unexpected players—including salt—that play key roles in autoimmune diseases.” They’re now taking the same approach with a wider variety of cells.
The wiring that network medicine reveals will be crucial, Regev says. “Without figuring out that circuitry, it’s hard to figure out what genes actually do and act on that knowledge.” And though such data is far from complete, Barabási says, “If you have an incomplete map of Boston, it’s still very useful. It can still tell you how to get from here to there. There may be easier ways to get there, but it’s still predictive.” Even an incomplete map, he says, may help improve medical care.
Finding the Network to Treat the Disease
Future medical treatments may not focus on a particular genetic mutation, but rather on the biological routes through which diseases are expressed. Such an approach could help in the case of certain melanomas, where patients with a particular genetic mutation were treated with a drug that inhibited that pathway. The melanomas receded, but within months, they returned full force. The cancer had found another pathway. By understanding melanoma’s network, doctors might employ many drugs that would act on many pathways, limiting its ability to evade treatment.
James Collins, a professor of biomedical engineering at Boston University, agrees that the diseases of the future will be managed with drugs that act on multiple pathways in the network. (He admits that pharmaceutical companies, searching for one drug to treat a disease, haven’t been particularly happy to hear that; developing, say, four drugs for a disease instead of one could be significantly more difficult and expensive.)
Another potential use of the diseasome could be in finding new uses for existing drugs. “Now we now what neighborhoods the diseases are in,” Barabási says, alluding to the genes and proteins that make up disease modules. “So, if a drug has been developed for heart disease but it tends to be hitting in the asthma neighborhood, it could be a good asthma drug.”
Though Barabási predicts that network medicine will offer solutions in the near future, Loscalzo believes the approach is already yielding results. In fact, Loscalzo created the Harvard Medical School Division of Network Medicine in 2013 to investigate diseases revealed through network medicine.
Working with the Barabási group, members of the Division of Network Medicine are trying to understand how changes in genes lead to disease—and to trace back through the network to find important genes that could be useful for treatment. They’ve been teasing apart the wiring of asthma. Loscalzo says they’ve used this approach to create subcategories of asthma. This could help determine, for instance, which patients might be more likely to benefit from specific treatments such as inhaled steroids, thus personalizing the asthma therapy.
Using a network approach has already led to one success today: in the area of antibiotic-resistance.
Loscalzo says they also identified a particular inflammatory pathway that led to pulmonary hypertension, where pressure in the arteries that carry blood from your heart to your lungs causes symptoms such as shortness of breath. The inflammatory pathway could serve as a potential drug target.
Using a network approach has already led to one success in a particularly thorny area today: antibiotic-resistant bacteria. Collins’ team tried to work out the process by which antibiotics killed bacteria. Scientists thought they already knew the answer, but Collins says they didn’t have the entire picture. He and his team’s research demonstrated that the drugs trigger stress responses in cells. Due to that stress, the bacteria themselves eventually create damaging molecules, which harm their DNA and contribute to their death.
But the chemicals that bacteria produce when they’re stressed do something else, too: they cause mutations in DNA that lead to antibiotic resistance. Today, antibiotic resistance is thought to emerge because, scientists have believed, there are a few bacteria in a given community that are naturally resistant to a drug, and they thrive after the drug kills off the bacteria’s brethren. But instead, as Collins’ research has demonstrated, antibiotics themselves induce mutations, leading to antibiotic-resistant bacteria.
With this new understanding, Collins and his lab set out to combat antibiotic resistance. Collins’ lab introduced an extra protein to antibiotics that flips on DNA repair activity within the cells. Repairing the DNA prevents the mutations that create antibiotic-resistance in the first place. In studies, this process boosted the efficacy of a drug such as Cipro from ten times to a thousand fold. Collins has since founded a company, EnBiotix, to attempt to commercialize this approach and improve the efficacy of existing antibiotics. He expects that the newly improved antibiotics—based on a network approach—could be tested in clinical trials within the next couple of years.
“Biology is complicated,” Collins says. The idea that scientists and drug companies can target all the physical expressions of a disease by going after just one gene is mostly wishful thinking, he notes. But he’s excited to be part of a growing community of scientists that are trying to understand all the pathways of a disease, among genes and proteins and the myriad other chemicals and signals in our bodies. And he’s “proud that network medicine is now leading to potential therapeutics.”
Barabási likens the recent history of medicine to the time when maps of the world were rare. Back then, we had no idea how people and places could be connected. But, he adds, “once we had maps, it was completely transformative for business.”
“In cell biology, there are no maps,” he says, “but they are starting to emerge.” And that could be completely transformative for medicine.
Image credit: Djebali et al. 2012, PLoS ONE