Visualizing Collaboration among Pharmas

One of the goals of the MIT Collaborative Initiative is to bring various stakeholders to the table and facilitate cooperation that will ultimately benefit the patients. But we are not the only people who practice this philosophy. The pharmaceutical and device companies have already realized the power of teamwork in clinical trials. Many companies work with each other in order to develop effective medicine and products. It is common to see a single drug sponsored by several Pharmas through different types of contracts and licenses.

Each partnership can be visualized as a link that connects two companies. With that in mind, it is possible to analyze collaboration through the use of networks. Using the BioMedTracker database, we were able to observe these collaboration networks using statistical and visualization tools. Given the advances of computational technology and mathematical graph theory, it is now possible for us to analyze the properties of these networks and develop insights on how effective drugs can be developed through research alliances. To give you a taste of the complexity, we started by looking at the top 100 companies with the largest market caps (Shown in figure 1).

Figure 1. Partnerships among top 100 companies with largest market capitalization . Red represents companies with more than 15 clinical trials collaborations.  Blue represents companies with less than 15 clinical trials collaborations.  Arrows are pointing from lead companies to partners.  Thickness of links represent the volume of partnerships. (Data from BioMedTracker)

It is evident that looking at a hundred companies create a web that is extremely complicated. It would be quite difficult to visually extract information from this network. Nevertheless, we can reduce this network down to 20 companies and rearrange the nodes to provide us a better visualization.

Figure 2. Partnerships among top 20 companies with largest market capitalization Figure 2. Partnerships among top 20 companies with largest market capitalization.

Although it is still complex, the preceding graph gives us a succinct view of the network. We can even tell that most of the large Pharmas such as Bristol-Myers Squibb, Sanofi, GlaxoSmithKline, Novartis, Pfizer, Merck, and Johnson & Johnson have large collaborations between each other. Another way to visualize this network is using a heatmap to color-code the number of partnerships between each company.

Figure 3. Heatmap of Partnerships .  Scale ranges from yellow to red where red indicates large volume of partnerships (~20) while yellow means no partnerships. The horizontal axis represents the lead company while the vertical axis represents the partner (Data from BioMedTracker)

On the bottom of the heatmap you will see a list of lead companies while the partners are listed vertically on the right hand side. From the first row of the heatmap, we can see that Bristol Myers and Sanofi frequently collaborate with Eli Lilly. This is useful for determining where interesting collaborations are.

Borrowing from the field of genomics, we can produce a circular layout of the network that is even more useful – and aesthetically pleasing. These types of visualizations are very common in the field of genomics for analyzing genetic information. This tool takes the same data as before and reproduces a figure that allows us to see the total volume of connections for each company as well as the connectivity.

Figure 4. Partnerships among top 16 companies with largest market capitalization The color of the links corresponds with the partner company (Data from BioMedTracker)

We can now see companies such as Johnson & Johnson have a large patch of its own color (neon green) coming from other companies. This means J&J tends to partner instead of being the lead company during clinical trials. On the contrary, Bristol-Myers Squibb (shown in violet) collaborates with several other Pharmas by being the lead company. This might indicate that Bristol-Myers expertise in leading and sponsoring clinical trials. However, deeper investigation must be done to determine if this is a fact.

Observing collaboration networks not only gives us insights into the nature of clinical trials but can be useful for determining where data, information, and knowledge are being transferred. These network maps and visualizations are only a starting point for network analysis. As a consequence of our hyper-connected world, network analysis is now relevant and will become an essential tool in the system researcher’s toolbox.

Gary Lin