This document discusses a biology research paper that uses a mouse model to understand the progression of Alzheimer's disease. It summarizes the paper's argument and evidence in four key claims:
1. Tau pathology is age-dependent in mice.
2. Human Tau is propagated synaptically in mice.
3. Mouse Tau is recruited by Human Tau to aggregate.
4. Tau pathology results in neurodegeneration accompanied by gliosis.
The summary provides an overview of the paper's central argument that tau and tangles lead to neurodegeneration, as supported by data from four experimental studies using the mouse model and citations of outside experiments.
3. Theoretical Framework
• Scientific paper as an argument
– “Scientific publications are documentary
representations of defeasible arguments,
supported by data and repeatable methods.”
(Clark, Ciccarese, Goble 2014).
Clark, Tim, Paolo N. Ciccarese, and Carole A. Goble. "Micropublications:
a semantic model for claims, evidence, arguments and annotations in
biomedical communications." Journal of biomedical semantics 5.1 (2014): 28.
4. Theoretical Framework
• Scientific paper as an argument
– “Scientific publications are documentary
representations of defeasible arguments,
supported by data and repeatable methods.”
(Clark, Ciccarese, Goble 2014)
• Micropublications model
Clark, Tim, Paolo N. Ciccarese, and Carole A. Goble. "Micropublications:
a semantic model for claims, evidence, arguments and annotations in
biomedical communications." Journal of biomedical semantics 5.1 (2014): 28.
5. Micropublications Model
[Clark, Ciccarese, Goble (2014) Micropublications: a semantic model for claims, evidence, arguments and
annotations in biomedical communications.]
5
6. Micropublications Model
[Clark, Ciccarese, Goble (2014) Micropublications: a semantic model for claims, evidence, arguments and
annotations in biomedical communications] 6
7. 4. Create a Micropublication
3. Identify the Reasoning
2. Annotate using http://hypothes.is
1. Identify a scientific research paper
Approach
9. Uh-Oh, It’s Complex!
Neurofibrillary tangles advance from layer
II of the entorhinal cortex (EC-II) toward
limbic and association cortices as
Alzheimer’s disease evolves. However, the
mechanism involved in this hierarchical
pattern of disease progression is unknown.
We describe a transgenic mouse model ….
10. Step 2: Annotate the Article
Conclusion
What we know about
Alzheimer's
What is
hypothesize
d to be true
11. Step 3: Identify the Reasoning
• Central Argument
– Tau & tangles lead to neurodegeneration
• Support
– Data from 4 experimental studies using a
mouse model.
– Citation of outside experiments.
13. Step 4: Create a micropublication
A mouse model of tau pathology is used to understand the
mechanism and progression of Alzheimer's disease (C0).
1. Tau pathology is age-dependent in mice (C1).
2. Human Tau is propagated synaptically in mice (C2).
3. Mouse Tau is recruited by Human Tau to aggregate (C3).
4. Tau pathology results in neurodegeneration which is
accompanied by gliosis (C4).
22. Citation of outside experiments
supporting claim 4
• Claim 4 is supported by the
biomarker materials, which are
justified by a chain of citations
– “PSD-95, a postsynaptic marker
that has been reported to decrease
early in neurodegeneration (Zhao
et al., 2006)”.
– Zhao in turn cites Lacor et al.
2004
23. Future Work
• Mine arguments by first extracting
methods from experimental research
articles.
• First: automatically extract methods and
manually infer the knowledge claims
that the method COULD support.
• Later: characterize the types of claims
authorized by the method.
27. Teufel, S. (2014). Scientific Argumentation Detection as Limited-domain Intention Recognition. In ArgNLP 2014
Frontiers and Connections between Argumentation Theory and Natural Language Processing. http://ceur-ws.org/Vol-
1341/paper14.pdf
Teufel’s model
28. Using Array tomography with two different
indicators red (post) PSD and green (pre)
synapsin I proteins to identify the densities of
pre and post synaptic structures.
29. [Clark, Ciccarese, Goble (2014) Micropublications: a semantic model for claims, evidence,
arguments and annotations in biomedical communications]
31. “[Y]ou can transform a fact into fiction or a fiction
into fact just by adding or subtracting references”
- Bruno Latour
Latour, Bruno. Science in action: How to follow scientists and engineers through society. Harvard University Press, 1987. p33
32. Funded grants with citation bias &
citation distortion.
Greenberg, Steven A. "How citation distortions create unfounded
authority: analysis of a citation network." BMJ 339 (2009): b2680.
https://doi.org/10.1136/bmj.b2680
33. Citing fake science harms people
A paper about a clinical trial for renal disease was retracted because:
“‘the trial had not been approved by the ethics committee, the involvement of
a statistician could not be verified, [and] the trial was not a double-blind study,
because Dr Nakao knew the treatment allocation’.”
“Nevertheless, the COOPERATE study was cited by 173 review articles and 58
secondary clinical studies that enrolled a total of 35,929 patients.”
“The harm done by COOPERATE is thus 4-fold:
• patients were enrolled in an experimental therapy for a condition which
already had an accepted therapy;
• time, energy and money were wasted by patients and investigators;
• false information pervaded the literature;
• and combination therapy was accepted more quickly and used more widely
than it might have been otherwise.”
Steen, R. G. (2011). Retractions in the medical literature: how many
patients are put at risk by flawed research?. Journal of Medical
Ethics, 37(11), 688-692.
34. Citing fake science harms people
A paper about a clinical trial for renal disease was retracted because:
“‘the trial had not been approved by the ethics committee, the involvement of
a statistician could not be verified, [and] the trial was not a double-blind study,
because Dr Nakao knew the treatment allocation’.”
“Nevertheless, the COOPERATE study was cited by 173 review articles and 58
secondary clinical studies that enrolled a total of 35,929 patients.”
“The harm done by COOPERATE is thus 4-fold:
• patients were enrolled in an experimental therapy for a condition which
already had an accepted therapy;
• time, energy and money were wasted by patients and investigators;
• false information pervaded the literature;
• and combination therapy was accepted more quickly and used more widely
than it might have been otherwise.”
Steen, R. G. (2011). Retractions in the medical literature: how many
patients are put at risk by flawed research?. Journal of Medical
Ethics, 37(11), 688-692.
35. Take away messages
• Biomedicine has evolved multiple approaches
for managing and appraising individual papers
and bodies of “facts”.
• Citations come in many shapes and sizes.
• Citations may support “facts” – as part of a
larger scientific fabric that includes data,
evidence, arguments.
• Powermove: identify the critical supports
(think Jenga)
36. Take away messages
• Network analysis may help identify
problematic citation practices.
• Modeling arguments can help identify the
robustness of a claimed “fact”.
• Semantic models could enable inference-
based reasoning and citation network
querying.
• Relevant citation corpora exist.
Editor's Notes
Argument visualization may help make research papers easier to understand, which could both speed quality assessment within a discipline and help build interdisciplinary knowledge networks. This paper presents a case study of the arguments in a single high-profile paper on Alzheimer's disease research. Within this one paper, we analyze and hand-annotate the main claim, which is supported by 4 sub-claims, in turn supported by data, methods, and materials. We also investigate how the paper imports and uses knowledge claims from other research papers. We create a specialized argument-based knowledge representation called a micropublication. In future work, we will investigate automatic argumentation mining for experimental biology research papers. Our long-term vision is to create literature-scale claim-argument networks that help more quickly use new knowledge about human health.
There are multiple levels of argumentation to study in science. Even when studying the argumentation of a single paper you can study it various levels. Some models focus on the macrolevel (for instance, Teufel’s model of the rhetorical moves and intentionality of a paper, with high-level goals such as showing novelty).
We focus on a deeper level of argumentation, identifying the knowledge claims and their support.
We use a model called micropublications.
You can think of the micropublications model as an argument map associated with a given claim. Both supports and challenges can be recorded in this map.
It is a scientific model of argument, developed for experimental, primarily biomedical fields. That gives the support and challenge a particular character.
A claim is supported by data that is generated by some method, which typically is supported by the material used. So for instance, an experiment can use mice, and the method might be to compare mice with and without some gene using some particular kind of equipment.
Overall, this is our approach. We annotate using a web tool called hypothesis.
The first step is to identify a paper. We use an experimental biology paper about Alzheimer’s disease, which is one of the most influential papers in the field. Its goal is to propose a mouse model for Alzheimer’s Disease.
As you can see from the first few sentences, this is a very technical paper. Novejot Sandhu, my student, was essential in reading and processing the paper.
Neurofibrillary tangles advance from layer II of the entorhinal cortex (EC-II) toward limbic and association cortices as Alzheimer’s disease evolves. However, the mechanism involved in this hierarchical pattern of disease progression is unknown. We describe a transgenic mouse model in which overexpression of human tau P301L is restricted to EC-II. Tau pathology progresses from EC transgene-expressing neurons to neurons without detectable transgene expression, first to EC neighboring cells, followed by propagation to neurons downstream in the synaptic circuit such as the dentate gyrus, CA fields of the hippocampus, and cingulate cortex. Human tau protein spreads to these regions and coaggregates with endogenous mouse tau. With age, synaptic degeneration occurs in the entorhinal target zone and EC neurons are lost. These data suggest that a sequence of progressive misfolding of tau proteins, circuit-based transfer to new cell populations, and deafferentation induced degeneration are part of a process of tau-induced neurodegeneration.
Over expression or abnormal phosphorylation (over activation of gene by adding a phosphate group) of tau gene results in a misfolded tau protein (PHF-tau) and neurofibrillary tangles which leads to neurodegeneration.
Conclusions based on 4 experiments on mouse models
Tau is transported from axons and accumulation of the protein is age dependent
There is tau protein in areas that do not express the human tau transgene which indicates that the pathology is spread downstream via neurons which was confirmed by FISH and staining
Misfolded human tau recruits the mouse tau into compartments which it is not typically expressed, the somatodendric compartment, where it becomes misfolded and contributes to the aggregation
Eventual axonal degeneration as a result of tau aggregates
We found 4 lines of argument supporting this claim, shown below as the four sub-claims C1, C2, C3, C4. The material (M), methods (m), and data (D) produce the sub-claims.
We found 4 lines of argument supporting this claim, shown below as the four sub-claims C1, C2, C3, C4. The material (M), methods (m), and data (D) produce the sub-claims.
The Data is in turn supported by Methods
Those Methods depend on the Materials used
Those Methods depend on the Materials used
Those Methods depend on the Materials used
6 figures, (36 subfigures, at least 5 subfigures each)1 A-E2 A-H3 A-G4 A-E5 A-E6 A-F
Today we search for authors, titles, and topics. Imagine being able to search for the method a paper used, its results, and the knowledge claims it can support.
This would help with a number of information retrieval problems.
It could also help us manage the
Could enable scalable knowledge representations
Enable argument retrieval (e.g.: data and methods)
Auditing argument networks
Summarized argument generation
This can make the role of research data and supporting evidence more clear. For instance, showing the dependencies between articles (Figure 2) could help identify possible logical inconsistencies in a network of scientific papers, and prioritize research for reproducibility experiments. Micropublications can also support more granular search and retrieval of aspects of scientific papers, such as the materials or methods used.
Micropublications can express connections of arguments between articles. For example, the evidence about Rapamycin is supported from different cited arguments.
How can I track and map research for a literature review?
“X.1 Being a Map of Physics
This map is the culmination of a six-year-long labor of love by noted physicist, visual artist, poet, and peace activist Bernard H. Porter. Porter began compiling the historical data upon which the map is based in 1932 while a fellow in radioactive research at Brown University. He then took most of the summer of 1933, working out of his parent’s home in Houlton, Maine, to compose the map’s striking visuals. The following years were spent circulating the map among prominent physicists and historians of science to verify its accuracy. The end result is a rich geography of a scientific field, one that uses mapping conventions to make understandable the way ideas move and develop over time. Ambitious in scope, the map traces the history of physics from the 6th century B.C. to the present day. Key theoretical starting points such as ‘Mechanics,’ ‘Sound,’ ‘and Light’ appear as water sources from which streams of thought emerge. Located alongside these rivers are “villages” representing figures like Isaac Newton, Alessandro Volta, Werner Heisenberg, and other major contributors to the development of physics. Surrounding it all is the map’s border, which is decorated with 24 diagrams that frequently appear in the work of physicists.
References:
Porter, Bernard. 1939. Being a Map of Physics. Courtesy of Maine State Library and Mark Melnicove. In "10th Iteration (2014): The Future of Science Mapping," Places & Spaces: Mapping Science, edited by Katy Börner and Samuel Mills. http://scimaps.org “
30 citations to the supportive data, 3 citations to the critical data. However, in a balanced network, all primary data should have approximately the same number of citations no matter whether it is supportive or critical data.
The COOPERATE study was a falsified clinical trial on 263 patients with non-diabetic renal disease.
The COOPERATE study was a falsified clinical trial on 263 patients with non-diabetic renal disease.
“As is commonly the case, different evidence is used by each lab - either because certain data were not accessible, or some labs judged certain data to be unreliable or irrelevant to the claim, or some labs interpreted the same data in different ways. SEPIO translates this scenario into the following narrative and set of instances to be represented in its formal modeling of the data.”
“As is commonly the case, different evidence is used by each lab - either because certain data were not accessible, or some labs judged certain data to be unreliable or irrelevant to the claim, or some labs interpreted the same data in different ways. SEPIO translates this scenario into the following narrative and set of instances to be represented in its formal modeling of the data.”