Big data and antibiotic research
Antibiotic discovery has now been enabled by Big Data. It appears that the algorithms used for Big Data may also help us lead healthier lives, and not just pinpoint advertising to us or take away our jobs
Over the last few weeks, I have had a few conversations with Anand Anandkumar, an old friend and boss of the firm Bugworks Research. Bugworks is a start-up drug discovery company that aims to discover novel pharmaceutical assets for combating antibiotic microbial resistance. Hans Christian Gram, a Danish scientist, is the first person to have discovered that bacteria come in two classes. The first, called Gram positive, is more susceptible to antibiotics, but the second, Gram negative, can develop resistance to antibiotics. Bugworks focuses on Gram negative bacteria.
The over-prescription of antibiotics, often to treat viral (and not bacterial) illnesses and their widespread use in poultry, fish and meat farms means that bacteria have had a chance to mutate and become resistant to many antibiotics. To add to this, we are using old ammunition. Anandkumar says that the last useful class of antibiotic now in current use, fluoroquinolone, was discovered in 1962—or over 55 years ago! This class includes drugs such as ciprofloxacin, which is used to treat a variety of infections, from Anthrax to infections of the gastro-intestinal tract.
Antibiotics that are used to treat the worst bacteria, such as the ones that cause deaths in hospitals when patients have initially come in for other treatments, still exist. They are used very sparingly, and only as a last resort. The irony here is that whether the patient lives or dies, he or she will not end up being a repeat user. As a result, drug companies pour very little of their resources into the discovery of such drugs, since the economic returns from finding a super class of antibiotics simply do not outweigh the costs.
Antibiotics have traditionally followed a pattern of research filled with discoveries via happy accidents. Mould (or a fungus) was used to come up with penicillin, the first antibiotic, and since then most antibiotic drug discoveries have relied on finding bugs that kill other bugs. For instance, bacitracin, which is found in Neosporin ointment, was first isolated from an infected wound of a child who had been hit by a truck. Some antibiotics have come from sewage pipes. Others have come from dirt. The pharmaceutical researchers managed to push the right microbes to grow in suitable conditions so that they could in turn be used to defeat humankind’s own bacterial enemies. As Anandkumar says, “Bacteria and other microbial organisms hate one another more than they hate humans so finding a friendly microbe that can kill a harmful microbe is the main method we have had so far. It is clumsy, but it has worked in the past.”
The explosion of ‘Big Data’ and sophisticated computing power to run complicated algorithms may eventually provide a light at the end of this tunnel. The use of this information technology (IT) can broadly be classified into three classes.
The first use of IT lies in the simple but effective mapping of disease patterns within a geography and various sets of people among its population.
This means that if a Kannadiga patient comes to see a doctor in Bengaluru with what appears to be a bacterial infection, the doctor can now use computing power to narrow down which specific bacterium may be involved, if any, and can therefore pinpoint his or her treatment much more effectively by prescribing the right antibiotic—or not prescribing an antibiotic at all if the illness is viral in nature.
The second use of IT is by using computational algorithms to presage a scientific molecular-level discovery, which is what Anandkumar and his team are attempting to do. They use advanced mathematical models and algorithms in their quest to design a chemistry-based solution that will allow for the isolation of certain proteins in the bacterium, which when disintegrated by a chemical substance, will kill the bacterium.
This sort of treatment attacks the proteins at various levels—the cell wall, the cell membranes or, as in Bugworks’s case, the Deoxyribonucleic Acid (or DNA) strands themselves, which contain the codes that help the bacterium replicate itself.
DNA can be thought of as a cluster of ribbons, all tied up together, says Anandkumar, and the permutations and combinations of how these ribbons show up in a cell run into truly vast numbers. Today’s enhanced computing power and the appropriate mathematical techniques can isolate which part of these ribbons need to be attacked, thereby denying the bacterium the ability to replicate itself. To be clear, Bugworks does not just use the mathematics of computational biology, it also uses a store of knowledge from microbiology, pharmacology and medicinal chemistry to then design the appropriate chemical compounds that can attack the resistant bacterium at a molecular level.
The third method is to use genomics, which essentially studies how the ‘enemy’, as Anandkumar likes to term harmful bacteria, has changed in response to antibiotic use, and then use this new-found knowledge to find out which changes in genomic sequencing need now to be attacked to kill the resistant bacterium.
The old technique used to combat antibiotic resistance was finding variants of known antibiotics, akin to the ‘accidental’ method I described above. Encouragingly, recent news says that a group of American and Russian computer scientists has created an antibiotic algorithm that can efficiently sort through databases—and discover 10 times more new variants than all previous efforts. The algorithm, called VarQuest, was described recently by Hosein Mohimani, a professor of computational biology at Carnegie Mellon University. He said in a press release that VarQuest completed a search that could have traditionally taken hundreds of years of computations. VarQuest’s search was made possible by the computing progress we have witnessed over the past few years. Antibiotic discovery has now been enabled by Big Data. It appears that the algorithms used for Big Data may also help us lead healthier lives, and not just pinpoint advertising to us or take away our jobs.
Siddharth Pai is a world-renowned technology consultant who has personally led over $20 billion in complex, first-of-a-kind outsourcing transactions.