An Algorithm To Predict Which Bacteria Are Good For You
For a more targeted microbiome fix than a fecal transplant
The gut microbiome—the trillions of bacteria that live in the intestines—is integral to health. Those microbes help digest food, affect immune systems, and can alter body weight or even mood. The problem is that microbiomes vary so much that scientists often don’t know just what has thrown it out of whack. That’s why some of the most hardcore have turned to fecal transplants, as a total microbial makeover.
Now researchers at Brigham and Women’s Hospital in Boston have developed an algorithm that could help predict how a specific microbiome will behave under a set of circumstances. And this predictive power could lead to new treatments for gastrointestinal and autoimmune diseases. The study outlining the algorithm was published in June in the journal Genome Biology.
Though every microbiome is a little different, in recent years scientists have started to understand how particular bacteria, in higher or lower concentrations, can affect a person’s health—the bacteria known as C. diff, for example, can cause diarrhea if it grows uncontrollably in the intestines. So scientists have started to explore the idea of “bugs as drugs”—instead of using a traditional drug compound, doctors could prescribe a cocktail of living bacteria to cure a disease caused by a microbial imbalance.
But because the microbiome is made of living organisms, it’s not easy to anticipate how it will react to something new. “With the microbiome, we’re dealing with an ecology,” Georg Gerber, the co-director of the Massachusetts Host-Microbiome Center at Brigham Women’s Hospital and one of the study authors, tells Vocativ. “You want [these therapeutic organisms] to stably colonize and grow in the host. You don’t want them doing bad things like causing good things to go away,” he adds.
Gerber’s vision is that someday a patient with a gastrointestinal or autoimmune issue could go to a doctor, give a fecal sample, and an algorithm like this one could help the doctor prescribe a probiotic cocktail suited to the patient’s specific needs to bring her microbiome back into balance.
This algorithm, called MDSINE, treats the microbiome like a dynamic community with behavior that can never be fully predicted. Unlike earlier, less sophistacited predictive algorithms, MDSINE determines the amount of uncertainty that a given therapy might work. “The major difference that we added is in terms of actually dealing with the noise and uncertainty in the data,” Gerber says. And, importantly, it can predict that dynamic behavior over time.
The researchers based their algorithm on the behavior of microbiomes tested in the lab. Then, to verify it, they checked it against the microbiome dynamics of five mice (dosed at a young age to have the same microbiome). The researchers altered the mice’s microbiomes so that the C. diff inside them grew out of control, giving them diarrhea and gastrointestinal distress. They then gave them an experimental cocktail intended to restore balance to the microbiome, taking frequent fecal samples over the 56-day experiment. Their algorithm predicted that the cocktail would work, and it did—“Having colonization by these [therapeutic] organisms has the desired effect,” Gerber says.
The human microbiome, however, is much more complicated than those in mice. So the researchers are now working on making their algorithm applicable to this increased complexity. That would certainly be useful for companies developing new bacterial cocktail treatments, since they could predict whether or not it would work before testing it, and doctors, too, who want to choose the right cocktail for their patients.
“Our hope is to be able to predict if a cocktail of these bugs in a patient will reach concentrations to be effective as a therapy,” Gerber says. “I feel very strongly that the point should be to get these tools out there.”