Biological knowledge discovery using the semantic web framework

标准的网络搜索引擎可能会找到描述生物化学物质的网页，但它们几乎没有能力找到具有特定属性的生物化学物质。其原因是，网络上的生化信息没有以机器可理解的格式表示，因此，不能以复杂的方式查询。如果生物化学是用一种正式的知识表示语言来表示的，比如描述逻辑，那么我们就可以用计算机来对信息进行推理，并推断出我们共享的概念化（即生化本体）中明确的知识。该建议旨在捕捉生化结构的语义，以便基于计算机的功能推断成为可能。该方法涉及生化本体的设计和推理能力知识库的开发。预期的结果是化学官能团可以从结构中推断出来，生化化合物可以自动分类。这个语义框架将为生物化学家提供在不同粒度的化学结构（电子、原子、分子）和功能（官能团、有机化合物）上进行查询的能力。这项研究有可能成为新出现的语义网的基石，从而以一种适合计算机推理的格式提供生化知识。我们预计，我们对化学结构的空间和时间方面建模的努力将有助于查询复杂的数据，如分子动力学模拟产生的数据。结合其他互补的努力，我们期望新的语义链接可以与我们的资源形成，从而改善数据集成和强大的新的数据挖掘机会，超越异构的生化知识。

Standard web search engines might find web pages that describe biochemicals, but they have little capability in finding biochemicals with a specific set of properties. The reason for this is that biochemical information on the web is not represented in a machine understandable format, and as such, cannot be queried in a complex manner. If biochemicals were represented using a formal knowledge representation language such as description logics, then we could use computers to reason about the information and infer knowledge that is explicit in our shared conceptualization, i.e. a biochemical ontology.This proposal aims to capture the semantics of biochemical structure such that computer-based inferences about function become possible. The approach involves the design of biochemical ontologies and the development of a reasoning-capable knowledge base. The expected outcomes are that chemical functional groups may be inferred from structure and biochemical compounds may be automatically classified. This semantic framework will provide biochemists with the ability to query at variable levels of granularity of chemical structure (electron, atom, molecule) and function (functional groups, organic compounds).This research has the potential to form a cornerstone of the newly emerging semantic web so as to provide biochemical knowledge in a format that is amenable to computer-based reasoning. We anticipate that our efforts to model spatial and temporal aspects of chemical structure will facilitate the querying of complex data such as that generated from molecular dynamics simulations.  In combination with other complimentary efforts, we expect that new semantic links may be formed with our resource, leading to improved data integration and powerful new data mining opportunities over heterogeneous biochemical knowledge.