monarch-initiative / SEPIO-ontology

Ontology for representing scientific evidence and provenance information
50 stars 11 forks source link

Characterization of the 'Evidence Line' concept #16

Open mbrush opened 7 years ago

mbrush commented 7 years ago

To date we have considered three perspectives on what an Evidence Line is and how it should be described in SEPIO: (1) as a collection of information; (2) as an argument ; and (3) as a reified relationship; and (4) an interpretation.

This ticket proposes definitions from each perspective, with the goal of settling on one as the primary perspective and definition for this concept. But we can still describe evidence lines form these other to further clarify and complement the primary definition.


1. Evidence Lines as Collections of Evidence Items

2. Evidence Lines as 'Arguments'

3. Evidence Lines as Reified Relationships

4. Evidence lines as Interpretations


Consider also the utility of the following descriptions and examples:

Description: An Evidence Line is created through the interpretation of one or more pieces of information that collectively support a meaningful argument for or against a proposition. To qualify as an Evidence Line, this argument must be independently significant as evidence - i.e. it must be capable of affecting the probability of accepting the target proposition as true. This does not mean, however, that it is independently sufficient to establish belief in the Proposition, as additional Evidence Lines may be required to ultimately accept the Proposition as true.

For example, in the ACMG framework establishes 'absence in population databases' as a type of Evidence Line that can argue for the pathogenicty of a particular variant. But this argument alone is not considered sufficient to establish a variant's pathogencity, as the other types of evidence are additionally required to establish the truth of this Proposition (e.g. a line of evidence demonstrating the variant to have a deleterious effect on protein function, or showing it to segregate with disease features in a family tree).

Example: An example of an evidence line would be the argument that a finding such as "Lepr1 KO mice exhibit lower blood glucose levels than matched WT controls" makes in support of the Proposition that "Lepr1 gene is involved in diabetes". The Evidence Items supporting this line of evidence could include experimental data from a study exploring blood glucose levels in Lpr1 KO mice, such as a 548.5 mg/dl measurement of blood glucose in a Leprtm1b/tmb1 mutant mouse, or a 1.3951e-24 p-value indicating this measure to be significantly different from wild-type mice.

Here, the finding and its supporting data exist independently of their use as evidence. An Evidence Line instance based on this finding comes into existence only when an agent interprets this it as an independently meaningful argument for a particular Proposition, in the act of making an Assertion.

A Note on Propositions: Recall that "Propositions" represent the sharable meaning expressed in a particular Assertion. They are abstract entities that, like numbers, are independent of space and time. They represent the core fact or meaning that is put forth as true in an Assertion. They are 'sharable' in the sense that the same Proposition can be expressed in separate Assertions made by different Agents, on different occasions. So a single Proposition that "Lepr receptor inactivation causes increased blood glucose levels" can be put forth in separate Assertions made by Dr. Smith and Dr. Jones.

Evidence is evaluated to assess the truth of a Proposition that is put forth in a particular assertion - but we will often talk about evidence as used to evaluate an Assertion as a succinct way to express this sentiment.