Linda Kato

Automated hypothesis generation based on mining scientific literature

Keeping up with the ever-expanding flow of data and publications is untenable and poses a fundamental bottleneck to scientific progress. Current search technologies typically find many relevant documents, but they do not extract and organize the information content of these documents or suggest new scientific hypotheses based on this organized content. We present an initial case study on KnIT, a prototype system that mines the information contained in the scientific literature, represents it explicitly in a queriable network, and then further reasons upon these data to generate novel and experimentally testable hypotheses. KnIT combines entity detection with neighbor-text feature analysis and with graph-based diffusion of information to identify potential new properties of entities that are strongly implied by existing relationships. We discuss a successful application of our approach that mines the published literature to identify new protein kinases that phosphorylate the protein tumor suppressor p53. Retrospective analysis demonstrates the accuracy of this approach and ongoing laboratory experiments suggest that kinases identified by our system may indeed phosphorylate p53. These results establish proof of principle for automated hypothesis generation and discovery based on text mining of the scientific literature.

SIMPLE: A Strategic Information Mining Platform for Licensing and Execution

Intellectual Properties (IP), such as patents and trademarks, are one of the most critical assets in today’s enterprises and research organizations. They represent the core innovation and differentiators of an organization. When leveraged effectively, they not only protect a business from its competition, but also generate significant opportunities in licensing, execution, long term research and innovation. In certain industries, e. g., Pharmaceutical industry, patents lead to multi-billion dollar revenue per year. In this paper, we present a holistic information mining solution, called SIMPLE, which mines large corpus of patents and scientific literature for insights. Unlike much prior work that deals with specific aspects of analytics, SIMPLE is an integrated and end-to-end IP analytics solution which addresses a wide range of challenges in patent analytics such as the data complexity, scale, and nomenclature issues. It encompasses techniques for patent data processing and modeling, analytics algorithms, web interface and web services for analytics service delivery and end-user interaction. We use real-world case studies to demonstrate the effectiveness of SIMPLE.

Predicting Future Scientific Discoveries Based on a Networked Analysis of the Past Literature

We present KnIT, the Knowledge Integration Toolkit, a system for accelerating scientific discovery and predicting previously unknown protein-protein interactions. Such predictions enrich biological research and are pertinent to drug discovery and the understanding of disease. Unlike a prior study, KnIT is now fully automated and demonstrably scalable. It extracts information from the scientific literature, automatically identifying direct and indirect references to protein interactions, which is knowledge that can be represented in network form. It then reasons over this network with techniques such as matrix factorization and graph diffusion to predict new, previously unknown interactions. The accuracy and scope of KnITs knowledge extractions are validated using comparisons to structured, manually curated data sources as well as by performing retrospective studies that predict subsequent literature discoveries using literature available prior to a given date. The KnIT methodology is a step towards automated hypothesis generation from text, with potential application to other scientific domains.

SIMPLE: Interactive Analytics on Patent Data

Intellectual Properties (IP), such as patents and trademarks, are one of the most critical assets in today’s enterprises and research organizations. They represent the core innovation and differentiators of an organization. When leveraged effectively, they not only protect freedom of action, but also generate significant opportunities in licensing, execution, long term research and innovation. In this paper, we expand upon a previous paper describing a solution called SIMPLE, which mines large corpus of patents and scientific literature for insights. In this paper we focus on the interactive analytics aspects of SIMPLE, which allow the analyst to explore large unstructured information collections containing mixed information in a dynamic way. We use real-world case studies to demonstrate the effectiveness of interactive analytics in SIMPLE.

Prospective Client Driven Technology Recommendation

Helping locate the patents of the right technologies for licensing to prospective clients is more than one billion USD business annually to IBM. However, searching for right technologies from multiple massive data sources for a value presentation to customers is a typical human labor intensive task in the past. In this paper, we design a prospective client driven technology recommendation system to enable the automatic search of technologies for patent licensing. The idea has been to make use of knowledges from the large-scale encyclopedia Wikipedia in conjunction with 11 millions patent documents to develop an online technology recommendation system for prospective clients of IBM. The live system demands not only the fast response time but also a set of highly relevant patent documents which are technically interesting to a query prospective client.

The Strategic IP Insight Platform (SIIP): A Foundation for Discovery

With billions at stake in new product development, acquisitions and alliances, IBMs Strategic IP insight Platform (SIIP) delivers transformative results, helping clients gain strategic insights. Applied to the pharmaceutical industry, SIIP accelerates the discovery of information to more quickly and accurately answer questions such as: which chemical compounds are good for which targets? Whats the likelihood this compound will succeed? What diseases could be treated with this target? What are the candidate drugs that can be re-purposed for a given disease? Applied to drug discovery in life sciences, the SIIP platform leverages and integrates a wide range of public and private content, rich set of deep analytics and a massive-scale architecture to improve patient outcomes. SIIP was born prior to the proliferation of the many big data tools available today. We describe what tools and architecture decisions have been helpful in this first phase of solution development, and what tools and architectures we are relying on as we raise our own standard for performance and service delivery.

Literature-based automated discovery of tumor suppressor p53 phosphorylation and inhibition by NEK2.

Significance We adapted natural language processing to the biological literature and demonstrated end-to-end automated knowledge discovery by exploring subtle word connections. General text mining scanned 21 million publication abstracts and selected a reliable 130,000 from which hypothesis generation algorithms predicted kinases not known to phosphorylate p53, but likely to do so. Six of these p53 kinase candidates passed experimental validation. Among them NEK2 was examined in depth and shown to repress p53 and promote cell division. This work demonstrates the possibility of integrating a vast corpora of written knowledge to compute valuable hypotheses that will often test true and fuel discovery. Scientific progress depends on formulating testable hypotheses informed by the literature. In many domains, however, this model is strained because the number of research papers exceeds human readability. Here, we developed computational assistance to analyze the biomedical literature by reading PubMed abstracts to suggest new hypotheses. The approach was tested experimentally on the tumor suppressor p53 by ranking its most likely kinases, based on all available abstracts. Many of the best-ranked kinases were found to bind and phosphorylate p53 (P value = 0.005), suggesting six likely p53 kinases so far. One of these, NEK2, was studied in detail. A known mitosis promoter, NEK2 was shown to phosphorylate p53 at Ser315 in vitro and in vivo and to functionally inhibit p53. These bona fide validations of text-based predictions of p53 phosphorylation, and the discovery of an inhibitory p53 kinase of pharmaceutical interest, suggest that automated reasoning using a large body of literature can generate valuable molecular hypotheses and has the potential to accelerate scientific discovery.

COBRA --- A Visualization Solution to Monitor and Analyze Consumer Generated Medias

Consumer Generated Medias (CGMs) --- such as blogs, news forums, message boards, and web pages --- are emerging as locations where consumers trade, discuss and influence each others purchasing patterns. Leveraging such CGMs to provide valuable insight into consumer opinions and trends is becoming increasingly attractive to corporations. This paper describes COBRA (COrporate Brand and Reputation Analysis), a visual analytics solution that surfaces the text mining and statistical analysis capabilities described in our earlier COBRA papers. Our interaction technique of search , visualization , and monitor enables detailed analysis of many CGMs without overwhelming the user. A suite of visualization solutions expose a variety of embedded COBRA visual analytics capabilities. Real world client engagements and user studies demonstrate the effectiveness of our approach.