29 Jul 2002 - AlphaDMax launch at DDT 2002, Boston, 5-9 August 2002

As researchers are all too aware, searching the scientific literature for relevant information is not only a tedious and time-consuming job, there is also a high risk that key information can be completely overlooked. AlphaDMaxTM now allows researchers to retrieve relevant information from large text sources more effectively. Based on Natural Language Understanding technology, this flexible tool bridges the gap between word and concept, between syntax and semantics. Domain specific vocabularies and ontologies will make this tool the ideal assistant for every life scientist.


Relevant data selection from text
Alpha DMax for text mining is designed as a digital research assistant. By means of query templates that are part of the tool, the researcher can specify items and relations of interest. AlphaDMaxTM subsequently selects all relevant articles from the vast amount of text material in databases (literature databases, experimental data, reports ...), and ranks them such as to best meet the researcher's expectations. Phrases and sentences that contain material corresponding to the researcher's queries are highlighted using a colour scheme. Ranking is based on the appearance of concepts of interest and relationships between them. AlphaDMaxTM extends traditional text mining by not only relying on word statistics, but, by looking for relevant semantic relationships. In other words, the emphasis is on looking for meaning in text. As a result, AlphaDMaxTM searches the literature just like a scientist would do, but significantly faster.

Application areas and background knowledge
In order to retrieve and extract information AlphaDMaxTM uses domain-specific background knowledge that has been provided by experts. Background knowledge can be added to the system as domain-specific vocabularies and ontologies, i.e. conceptual schemes that indicate how the items in the vocabularies relate to one another. Current applications focus on protein-protein interactions, protein function, gene expression profiles and drug action. AlphaDMaxTM is therefore optimised for use in areas such as functional genomics, proteomics, chemical genomics, Structure Activity Relationship analysis (SAR) and toxicogenomics. New or additional background knowledge can also be added to the system, widening its potential use to many other drug discovery tasks.

A reliable, flexible research assistant
AlphaDMaxTM attributes a numerical confidence to all extracted data, based on source value and the frequency of targeted concept and relational occurences. This confidence indicator is an important parameter that enables scientists to judge the value of the obtained data and patterns. By providing substantiated and balanced information, the tool offers the researcher explanations in addition to results. These may provide a stimulus for the extension and enhancement of knowledge in a special domain of interest. AlphaDMaxTM is a user-friendly, interactive research tool and a flexible assistant that can be consulted at any time and at all stages of research.

Structured patterns for efficient analysis
AlphaDMaxTM opens relevant data in the literature for automated analysis by converting the data from free-style text format into well-structured computer patterns (templates) that are stored in an integrated database. Graphs, diagrams and pictures can also be stored in the database and accessed using image analysis technology developed by PharmaDM. This more complete set of information can subsequently be analysed using DMaxTM, PharmaDM?s data mining suite. AlphaDMaxTM is therefore an addition to the rich pool of data mining solutions offered by PharmaDM.

PharmaDM technology allows for mining an integrated database of biological, chemical and clinical data sources (numerical, symbolic, graphical, images and scientific literature) in search for significant and meaningful patterns. Using PharmaDM?s unique second generation (relational) data mining technology, the database can be searched directly from a variety of viewpoints, including data from different and diverse fields of expertise. Most significantly, relational data mining removes the risk associated with first generation data mining products where data reduction is an essential but harmful step.

PharmaDM is a Belgian based spin-off company from the universities of Leuven (Belgium), Aberystwyth (UK) and Oxford (UK). Although recently founded (15 November 2000), the company builds on expertise and technology accumulated at these institutes over the past 10 years. PharmaDM develops Integrated Data Mining Solutions for drug discovery research.