Published September 14, 2017
Buffalo has had an “outsized” influence in the field of biomedical ontology over the past 15 years or so
That's thanks in large part to the work of UB’s Barry Smith, PhD, Julian Park Distinguished Professor of Philosophy; adjunct professor of Biomedical Informatics, Computer Science and Neurology; and director of the National Center for Ontological Research. So says Alexander Diehl, PhD, associate professor in the Department of Biomedical Informatics in the Jacobs School of Medicine and Biomedical Sciences, who, along with Smith, will be among those representing UB at the 6th Annual Workshop of the Clinical and Translational Science Ontology Group at the University of Michigan on October 25 and 26. According to Diehl, Smith has been “an incredible contributor to and a leader in the field.”
Smith inaugurated the now-annual International Conference of Biomedical Ontology in Buffalo in 2009. The stated goal of the conference is “to explore new and existing uses of common ontologies to support creation, sharing and analysis of clinical data.”
Ontology, says Diehl, “was originally a philosophical discipline having to do with the enumeration or naming of things that exist in the world.” Applied ontology means using real-world categories to organize and analyze empirical data, and, through AI, to discover patterns and connections in those data that could potentially lead to practical medical advances and new therapeutics.
“What we’re really doing is trying to name entities that exist in the real world — things like objects, people or cells, but also things like processes, like walking or metabolism is a process — and we want to describe these things both in terms of what actually exists but also how they are related to each other,” he said.
In anatomy, for instance, “we talk about different parts of the brain, which parts are adjacent to other parts, how connected they are to each other,” Diel said. The main body of a neuron may lie in one layer of the brain, but its dendrites and axons may extend to other portions. Beyond a mere listing of parts, “we want to be able to describe that cell, exactly how it’s connected in the brain,” and one way to capture those connections is through applied ontology.
“The advantage of encoding all these logical connections between different entities is that then computers can use these logical statements to do inference. Say, for instance, that you have experimental evidence that a particular gene is expressed in a class of cells, a class of neurons, and then you know the parts of the brain that these neurons are found in. That gives you information about where the gene might be expressed within the brain as a whole.”
Most medical disciplines already have standards committees which sanction definitions of terms. Naming, however, is not the ultimate goal of the ontologies. “We want to use standardized names, but what we really want is interoperability between ontologies,” said Diehl.
In his own experience as an experimental immunologist in UB’s Department of Neurology, Diehl worked on ontology projects related to neurology and infectious disease, and continued his long-term work on the Cell and Gene ontologies. Currently, he is working with colleagues at Roswell Park Cancer Institute and other institutions to build out cancer ontologies for analyzing cancer data. Specifically, he plans to form a working group to develop a cancer microbiome ontology.
“All the ontologies we make are intended to be part of a larger project called the Obo (Open Biomedical Ontologies) Foundry,” Diehl said, a project that Barry Smith was instrumental in launching.
“The Obo Foundry organization specifies certain principles we follow for ontology development,” said Diehl. “The goal is to make sure these ontologies are interoperable, meaning that, for instance, when we talk about a particular type of T cell in the Gene Ontology there’s an understanding that they accept the way it’s been defined in the Cell Ontology, because the Cell Ontology is the main ontology for that domain.”
The work is inherently cross-disciplinary and collaborative, in addition to being data-driven. “The Gene Ontology Consortium is based in multiple institutions both in the U.S. and the U.K.,” Diehl says.
Diehl’s hope is that using ontologies will “enhance cancer treatment through better data analysis, both in the sense of speeding up the way we analyze how particular lymphomas and leukemias are progressing but also by improving the analysis of even older data, after it’s been collected, by looking for new patterns which the ontologies help to spot.”
Diehl’s decision to assemble a working group in cancer microbiome ontology was informed by recent trends in the field. There has been growing interest in the microbiome due to its ability to modulate the immune system and, more directly, due its role in the development of stomach cancer.
“We’re trying to model reality,” he said, “but our understanding of reality through scientific methods changes over time as we get new data.”