ConceptualEyes for Medical Research: A Showcase

Literature Survey

Conduct literature surveys in hours versus months

Drug Discovery

Evaluate drug targets and small molecule pathways rapidly

Hypothesis Generation

Assess near-infinite number of possible associations to innovate

Expertise Augmentation

Enable experts to do their best by utilizing what computers do best

Proven value proposition in medicine...

1

Health and medical innovations are facing a Big Data problem

Today, a biomedical/pharmaceutical researcher needs to peruse and digest 25000+ titles, 2500+ abstracts and 250+ research papers to design a novel hypothesis or experiment. This process would take 6-8 months for an experienced academic who is faced with 2 million new research articles published every year.

2

The Big Data problem is much bigger

 Over 6000 diseases do not have a cure. To discover, develop and launch a new drug, it costs over a billion (U.S dollars) and over a decade of work. Meanwhile, hundreds of millions of people suffer across the world in desperate need of rapid diagnosis and effective affordable therapies. 

 

3

ConceptualEyes empowers researchers

ConceptualEyes shortens the process of discovery from months to hours by creating searchable "Knowledge Universes" for medical innovation - empowering researchers to explore the what-is, connect the what-if, conceptualize the what-else and hypothesize the what-could-be.

4

Intelligence augmentation using artificial intelligence

ConceptualEyes hosted a knowledge-graph of medical facts extracted from millions of articles in the medical literature starting 1865 that was made available by the National Library of Medicine. It served as an intelligence-augmentation platform that applies semantic, statistical and logical reasoning on natural language to interact with domain-specific knowledge.

5

Success stories 

Our tools have already been used to design workflows for mystery-illness diagnosis, rare-disease discovery, association of diabetic retinopathy and beta-blocker treatment of hypertension and exposition of compounds such as xylene as an environmental carcinogenic risk. Currently, the tool is being applied in literature summarization around specific areas of research – such as schizophrenia and autism, discovery of new nutraceuticals, protein pathway storyboarding, rare-disease diagnosis and cures.

Case Study: Rare-disease hypothesis generation

Problem: 

Hypothesize cause-of-death of historical figures (e.g. Oliver Cromwell, Mozart, etc.) given medically relevant anecdotes of their life from historical biographies.

 

What ConceptualEyes Did: 

ConceptualEyes, then called ORIGAMI, used its natural language processing, graph analytic and artificial intelligence capabilities to hypothesize the cause of death. It used publicly available medical knowledge obtained from PubMed and reasoned with knowledge facts to generate hypothesis. Compared to other symptom-based diagnostic tools, ConceptualEyes not only produced the hypothesis, but also the provenance explaining the hypothesis through many years of published academic research. 

 

Results:  

In a few seconds, ConceptualEyes converged on the same conclusion reached by medical experts that had deliberated for weeks. When the results were compared to an independent crowdsourcing experiment with a panel of doctors, ConceptualEyes identified the consensus answer and generated 60% of the hypothesis doctors did, in two consecutive years - 2015 and 2016. 

 

Read more:

Los Angeles Times, January 2016 

Case Study: Beta-Blocker Treatment and Diabetic Retinopathy

Problem:

Subject-matter-expert provided a dataset of diabetic patients that was collected for building a predictive model of retinopathy-risk and poses the question - What are the patterns in the data that could augment or dispute observations in the literature? 

What ConceptualEyes Did:  

Integrated knowledge-facts and domain-specific data, and using its AI core analyzed 7600 diabetic patient’s 31 different clinical lab measurements with over 100 meta-variables collected over 6 years. The analysis revealed patterns of significance that were presented to the subject matter expert for interpretation.

Results:

ConceptualEyes identified two cohorts: (i) patients with a diagnosis of hypertension with Lisinopril as the treatment drug, and (ii) patients at risk of retinopathy with a pre-existing description of fluid and/or exudates in the eye exam and those taking insulin along with oral medication for blood sugar control. ConceptualEyes automatically revealed the association between the medication for anti-hypertensive treatment and retinopathy-risk, particularly the ones using β-blockers as treatment and generated a new question that needed exploration.

New Research Question:

Do β-blockers for hypertensive treatment accelerate severity of retinopathy in diabetic patients?

Read More:

Proceedings of IEEE International Conference on Big Data, 2016

User Testimonials

1

Dr. Plavi Mittal, Jain Foundation

“It would be fun to go back to graduate school with this tool at my finger tips…”.

2

Dr. Ryan Yates, Seattle Discovery LLC

“My forays produced more questions than answers and the answers are pretty darn awesome. The tool was able to meaningfully connect a target and molecule. The same terms on PubMed or Google search returned nothing….”

4

Dr. Georgia Tourassi, Oak Ridge National Lab

“When we threw the EPA’s top 10 carcinogens, we noticed that there were a few elements that appeared over and over as connecting links. Some of these elements made sense from a reasoning point of view, but there was one that we had never seen before...”

Interested in a demo?