Having an understanding of how the world ticks is at the core of developing our opinions, hypotheses and ultimately making decisions. When we see dark clouds hanging in the sky, we know that it will likely rain and we can choose to take an umbrella with us. This understanding of causality applies to our daily lives but also in highly specialized domains such as Biomedicine.
One of the things that biomedical experts are interested in, is to understand how diseases interact with other diseases and which proteins play a role in the development of a disorder. This causal understanding of biomedical mechanisms is key for developing treatments or diagnosing patient conditions.
However, there are more than 28,000,000 million academic publications listed on PubMed containing information on the intricate relationships of biological systems. On top of that, the number of publications is increasing exponentially with PubMed currently adding more than 1,000,000 citations per year.
How can we cope with this information volume and still make sense out of the data?
Today, if we like to know “what substances can cause liver cancer?”, we typically follow three steps:
- Find and read the documents that talk about liver cancer and different chemical agents
- Connect the dots & document insights
- Form our opinion / hypothesis and make a decision
Looking at the steps above, we spend most of our time reading documents and connecting the dots as we are collecting evidence to answer our question. Besides the time effort, this is an error prone process due to reading fatigue and selective reading bias. Our best alternative today is to read only a fraction of the available information e.g. 100 of 10,000 documents and document our insights as good as we can.
What if somebody had already read all the documents, had connected all the dots and recorded all relationships ? And what if we could have intuitive access to causal models based on evidence from thousands of documents ?
The benefits would go far beyond saving time – experts could discover connections that they hadn’t thought of before, and models could be used to make predictions (stay tuned for a blog post on this!). Decision-makers and analysts in Industry could quickly generate big-picture insights such as new emerging causal knowledge in interesting disease areas, or visualizing disease networks.
At Causaly we are developing a machine-reading platform that processes millions of documents and connects the dots into causal knowledge graphs. We then enable users to explore causal associations and gather evidence for their hypotheses instantly. This information can be visualized and explored intuitively.
Apart from semantic search, researchers can make use of advanced analytics capabilities for network analysis and entity-link predictions to arrive at novel hypotheses, using the combined knowledge from millions of documents.
We are excited to start working in the Biomedical domain and to empower scientists and decision-makers in Pharma and Academia. Building a causal model of the world of course entails more than the biomedical sciences and over time, we aim to expand into other verticals and make its applications available to our users. It’s a long road, but we’re looking forward to a good (long) journey.