Navigating through the open waters of published medical articles is a difficult proposition. Millions of articles, complicated keywords (often spelled differently, or with arcane abbreviations and acronyms), thousands of publications, all that makes a web of knowledge virtually impossible to mine and use effectively at the point of care. Even such powerful tools as PubMed itself often return tens of thousands of results for common keywords. CureHunter, Inc., a Portland, Oregon company, plans to change all that with its online automated tool to perform meta-analysis of the medical literature in real time.
Medgadget had a chance to talk to Judge Schonfeld, CEO of CureHunter, and a main architect of the company’s artificial intelligence-based medical text mining system, called CureHunter Discovery Engine.
Schonfeld is an executive and scientist with a 40-year career in applied computational linguistics (Google search) , formerly with UCLA’s Semantic Foundations Team. He describes CureHunter not just as a search engine that returns all the possible results, but rather as an intelligent language-based medical text mining system developed to read the entire Medline archive of the US National Library of Medicine, to automatically extract the key evidence of successful clinical outcomes and present it in easy to read format.
Schonfeld gave us an example of Amantadine, a drug with numerous synonymous names, just some of which are Aman; Symmetrel; tregor; Symadine; and 1-Aminoadamantane.
Ordinary search engines often miss multiple names and applications in research for the same agents and diseases when they are spelled differently and thus their “counts of relevance” are often totally misleading.
They also miss the fact that numerous drugs are used for treatment of a great variety of diseases, from Parkinson to influenza in Amantadine’s case; and ordinary engines have no way to find and structure those relationships for added understanding and prediction of outcome results.
CureHunter by contrast finds and corrects for all variants of drug and disease nomenclature to correctly group related properties. But most importantly the system reads each sentence in the literature base to see if a very specific statement of effective clinical outcome is reported for the drug and disease relationship. The strength of the outcome is then quantified by the system for predictive calculation of relative drug merit.
But what if the drug’s multiple usages were never recognized together or directly linked for analysis?
We might well miss many new target applications for good medications and more importantly miss common properties and causative agents of related diseases.
In Schonfeld’s words, numerous drugs or novel compounds, have similar re-targeting potential, but those potentials are likely buried in a haystack of medical knowledge, and CureHunter’s data mining technology can help to discover them automatically.
CureHunter’s interface is neither cluttered nor hard to learn for such a complex and powerful tool. All key evidence can be accessed from the same page from the one word entry of a target drug or disease name.
Drug performance graphs and history charts are produced automatically,typically in 10-20 seconds and thus are useful in real time.
Schonfeld has provided the following details on the company’s function and services available at CureHunter.com (please note that for now, for the duration of beta testing, and maybe beyond, the site is free, and the professional research interface is open to all licensed MDs):
We have divided CureHunter product applications (and markets) into 3 general areas:
1. Patient-Physician Summary Reports are really disease-specificmonographs: They are summaries of the clinical outcomes found in all peer-reviewed research articles only and allow patients and physicians to see at a glance the most well-documented drugs to treat diseases. Unlike ordinary health info sites which provide generic information, CureHunter Patient-Physician Summary Reports actually rate the best medications based on the sum of the scientific evidence for their usage. Physicians, especially in General or Family practice can come rapidly up to speed on the best meds for a broad range of conditions by reading a report–which are comprehensive 1949 – 2007 > and updated daily.
If distributed by a cooperating patient association, 50% of all revenues go to the patient group to support their own efforts and research. All interested associations are invited to contact the company.
2. The CureHunter On line evidence system for Physician Clinical Decision Support is accessible now, from the company’s home page, www.curehunter.com for free by all licensed MDs. It enables doctors to see at a glance an instant charting of the medications with the most supporting evidence for their use. A few more clicks and sources of the evidence are displayed in line on the same screen.
The CureHunter system is very unlike common professional interfaces that work like Google (or dumb terminals) and ask the doctor to write queries, vary spellings, and generally guess at relevance of the articles returned and go off and read them all to be sure of any merit. All results in CureHunter are precisely relevant to start with and used in the computation of the drug performance graph. They function as citation references for every data point in every graph.
Search engine technology is not comparable at all. CureHunter is an intelligent system that understands a lot about what it is reading.
Technologies in the System Include: NLP or Natural Language Processors, Computational Linguistics, Computational Biology, Artificial Intelligence, Network Graph Theory and Ontology or Knowledge Systems Theory.
The graph of relative drug performance merits is extremely useful when patients are not responding to first and second line default medications and the physician must proceed to an “educated guess”:
About 30% of all patients seen in clinic for the first time. That “guess” will be much more educated and vidence-based if mapped on a CureHunter graph. Each graph is essentially a real time meta-analysis.
The online interface can be piped directly into the leading EMRs, EPIC, GE Centricity, VA CPRS, and NextGen so that checking the evidence is a normal part of clinical care and record keeping. This instant availability in the record allows doctors to always have documented evidence support for their drug decisions.
The company believes an SOP, regular evidence check at patient reception would: Improve positive outcome rates, minimize adverse events, increase overall safety, and reduce costs dramatically by minimizing recurring visits and tests to change ineffective or patient intolerable medications. Large HMO payer-providers stand to save $500K dollars for every 1,000 patient visits.
3. Pharmaceutical research for drug discovery and development based on Network Graph Theory.
The CureHunter system automatically seeks the drugs and diseases that cluster by virtue of sharing key biological agents and mechanisms of action that have achieved clinical success. This function allows computation of new off-label uses for existing medications and provides insight into many disease mechanisms that previously would be hidden in mountains of data spread across millions of articles and decades of research.
For the pharmaceutical manufacturer CureHunter’s ability to predict new uses for medications or molecules in its portfolio can have a dramatic improvement on pipeline fill and profitability.
CureHunter’s method for new drug utility prediction is unique. If a doctor enters a disease name and then clicks the Network Graph Tab in the main interface, related diseases are graphed automatically to reveal syndromes and co-morbid illnesses that are commonly seen, and may share a single effective medication or key molecule. For example, psoriasis and arthritis may both be treated with methotrexate, cyclosporin, or one of the newer biologics like Enbrel or Remicade. The possibility that a medication can be cross effective is computable in a matter of hours from the CureHunter master data set of significant outcome evidence that can cross check millions of agents for common functionality in thousands of diseases.
Primary pharma manufacturers and research partners are invited to contract with CureHunter for portfolio analysis.
Schonfeld tells Medgadget that CureHunter is already sharing data with a number of universities, research institutions, and pharma companies to assess the usefulness of its technology, to advance both pure research, and to improve patient care. Lastly, the upcoming National Science Foundation’s Biomedical Informatics Workshop (December 4th & 5th, 2007) will also take a look at CureHunter’s technology.
More at CureHunter.com…
Press release: CUREHUNTER Deploys “Machine 2nd Opinion”- Research Interface for Doctors ..