Science - Network medicine

Mapping the human interactome

May 24, 2017

Reading Human Genetics, by A.M. Winchester, in high school was an inspiration because it revealed to me the hidden, intricate machinery of genes within our bodies. I was fascinated by how one gene mutation could lead to very complex diseases and syndromes. Later in my studies, I wandered through mazes of genes, and discovered how disruptions in the signaling proteins coded by these genes affect cellular function. This journey has brought me to Biogen, where I work with a talented team that is mapping the human interactome, the complex protein network and its thousands of interactions, in order to create better therapeutics.

Protein networks
This is easier said than done. Navigating a protein network is like following a thread through a social network. It’s endless, with twists and turns, unexpected players and unforeseen treasures, and it’s like trying to determine who is affected by what post. When you “friend” someone, you may end up with an instantaneous new group of friends; post a selfie, and suddenly, your 3rd cousin, twice-removed may want a reunion.

Protein networks are similar: one disease-causing mutation directly alters the functions of that protein but its effects go much further. This makes the identification of a good drug target time-consuming and difficult. But as you can get a pretty good idea about a person from their social network (friends, family, posts, likes, etc.), finding out a protein’s interacting partners is the first step towards figuring out its function and role in disease. Fortunately, like people, proteins with similar function, including those implicated in a disease, usually hang out together in the cell.

The source code
We’ve set out to make an unbiased blueprint of all proteins and their interactions, using individual protein purification and mass spectrometry. This is not an easy feat, as there are about 20,000 different proteins in humans, with about 100,000 variations. I was part of a large-scale effort that generated the most comprehensive map of protein interactions for fruit fly cells, published in 2011 in Cell, which has a much simpler proteome compared to humans.

Spyros Artavanis-Tsakonas, then Chief Scientific Officer of Biogen, championed the importance of such studies in drug discovery and in 2012, we initiated the Human Interactome Project in collaboration with Harvard Medical School. The project was co-funded by the NIH. Like the name suggests, we are defining the cellular network of protein interactions. We’ve completed Phase I, and created a map of protein complexes of the normal (baseline) human proteome.

Together with Steve Gygi and Wade Harper at Harvard Medical School, we analyzed 10,000 proteins in a human cell line. We expressed one tagged protein at a time, and asked what these proteins do, where they reside, and what other proteins they bind to. We have scored millions of combinations and have learned how to weed out false-positives as well as zero in on true interactions. In other words, we determined which proteins are true “friends” and which ones are free-loaders - those proteins that are there by coincidence and have nothing to do with function. Our initial results were published in 2015 in Cell, where we report on 23,744 interactions, of which, 86% were previously unknown. This map also provided the first experimental evidence of molecular function for many human proteins.

In May 2017, we published in Nature an expanded map, called the BioPlex 2.0, identifying protein interaction partners for more than 5,800 protein-coding genes. We have now mapped interactions for more than a quarter of the human genome. The updated network comprises over 56,000 unique protein-to-protein interactions among nearly 11,000 proteins. It expands the protein complexes associated with fundamental cellular processes and connects many genes implicated in human diseases.

Data-driven research
How do we hope to use this information to generate better drug targets? Our approach is holistic and data-driven: we don’t start with a specific hypothesis, which seems counter-intuitive. Rather, we look at networks agnostically and let the data tell us what genes/proteins mean to each other (or not). We are beginning to understand how information flows in these pathways/networks. Using this approach, we can integrate disparate datasets to identify previously missed connections, enhance the investigation of difficult-to-target diseases and even predict which proteins are compromised in a disease. A better understanding of the molecular surroundings of a disease-causing protein in the cell can help uncover additional drug targets or reveal useful biomarkers.

Let me explain. Several disease-related genetic mutations have been identified in ALS. These ALS genes don’t have much in common, except that some are involved in RNA processing. Using fly genetic screens, our group, including Mark Kankel and Anindya Sen, also identified several other genes that can alter the disease severity or progression. It was not clear why these genes were important for ALS, or if they would be good targets. After interrogating our interactome and other knowledge bases, we hypothesized that there are distinct paths and potential convergence points in the network that might help explain why all these mutations in different genes lead to the same disease in patients. Long story short, we have discovered a handful of genes that appear to suppress four major genetic forms of ALS. These appear to be very promising targets for all ALS, including perhaps sporadic ALS cases.

Network medicine
Until recently, our pursuit of new medicines has been based on the reductionist model of targeting a single genetic defect to design an all-in-one therapy. Network medicine is the opposite: it investigates the interdependencies and plethora of influences that disturb the system in a disease, and the possibility that these factors may also be associated with other adjacent disorders. The challenge is in uncovering and targeting critical points in the network to alter the effects of the underlying mutation/disease. For me, the interactome provides the molecular framework upon which one can layer large-scale genomics, genetics, expression profiles, big data and other disease-relevant datasets with the aim of expertly navigating and identifying unique drug targets for complex diseases.

Guruharsha Kuthethur Gururaj, Senior Scientist, Neurology

Related Perspectives