PATO meeting with Hymenoptera AO and Phenoscape
25-27 September 2010, Eugene, Oregon
Ontology and phenotype representation issues
Attending: James Balhoff, Andrew R. Deans, George Gkoutos, Eva Huala, Suzanna Lewis, Paula Mabee, Chris Mungall, Katja Seltmann, Monte Westerfield, Matthew Yoder
Legend: Action Items Covered Topics CompletedAction Items
25 September 1:00 pm - 6:00 pm
I. PATO questions and issues:
1. Jim: Gradational size phenotypes in which states include small, medium, large. In comparisons across species, size is often described along a continuum of, for example, “small, medium, large”. PATO only allows for small (‘decreased size’ PATO:0000587) vs. large (‘increased size’ PATO:0000586) comparisons. Only these two terms are currently in PATO vs. others such as ‘medium’ or ‘very large’ (add?). How to describe continuum? No solutions proposed to date.
Maybe not a problem going forward (if systematists change practice....). But problem for legacy characters. Add future features that warn users if they make size-based statements that
More than one problem (Suzi) -- might have three, ten, bins, etc. People won’t ever choose the same number of bins. What is it all relative to?
Action item: Chris & Jim will create a formalism for doing this - and Jim must build an implementation in Phenex (Jim needs to provide this - impt for Phenoscape group mainly).
a. create ordered bin/partitions that are relevant to the study (via Phenex interface?). Every member of these set of bins would be relative-to a particular entity (e.g. the fin of this taxa). These bins would be sub-classes of size. (per study “vectors”) could further annotate or relate these vectors with absolute size or “real” size ranges which might allow for additional more specific queries (if these values exist).
Example (option 1):
* bins pf1-5 in study A describes ‘pectoral fin’
* bins pf1-5 in study A describes size
* bin-pf1-studyA < bin-pf2-studyA
* bin-pf2-studyA < bin-pf3-studyA
* ...
* bin-pf4-studyA < bin-pf5-studyA
* taxon t1 exhibits bin-pf1
* ...
b. Using relations “increased in magnitude relative to”, “decreased in magnitude relative to”
Entity->new(size bin)->working in application->populate members in bin as observed
Every phenotypic statement is a comparison, i.e. a relative statement, even if the norm it is being compared to is not explicitly stated. The issue is not with PATO per se, but with representing this in the statements. Phenoscape represents this coarsely by annotation to “size”.
Two issues: composite qualities, and a subset- composite qualities that are composed of terms from two ontologies.
Comes down to a application issue, with only relative changes comparable across taxa shared in studies. Study(size_bin(taxa)).
2. Paula: Position vs. Relational spatial qualities.
•Position: (is_a physical quality)
– A spatial quality inhering in a bearer by virtue of the bearer's spatial location relative to other objects in the vicinity.
•Relational spatial quality (is_a quality of related physical entities) (maybe change string to “relative position”)
–A quality of related physical entities inhering in a bearer by virtue of the bearer's position pertaining to or involving or having the nature of space in relation to another entity.
a.: When to use relational spatial vs position for phenotypes involving two entities?
e.g., Mental foramen: 0, positioned just anterior to coronoid process; 1, forward of midpoint of the gnathic ramus of dentary Corburn and Cavender 1992, #81:
•EQ statement 1: E: mental foramen; Q: anterior to; E2: coronoid process [relational spatial quality]
•EQ statement 2: E: mental foramen; Q: position^anterior_to (coronoid process) [post-comp with positional quality - analogous to size] Chris says that this is ill-formed because we are using anterior to as a relation between a quality and an object. DITCH THIS.
Action item: (Chris) Look at what is under position. Might remove position after considering subclasses. [subsumed under # 14]
–Example “frontal bone larger than parietal bone”
–E: frontal bone + Q:size^increased_in_magnitude_relative_to(parietal bone)
b: Angular placement / Orientation of a structure:
e.g. ‘Coronoid process of dentary: 0, directed posteriorly; 1, directed dorsally (Mayden 1992, #124)
EQ statement option 1: coronoid process Q: orientation has_quality (posterior)
EQ statement option 2: Coronoid process Q: posteriorly rotated [child of position]
Two questions:
1. How to annotate characters that describe what direction a projection is pointing? For example, “process points anteriorly”. Orientation doesn’t seem like the right quality (see def) to describe what direction the process is pointing. New quality such as below:
Def of orientation =angular placement: A relational spatial quality inhering in a bearer by virtue of the bearer's placement which is defined by the angle between the bearer and an axis, or the angle between the bearer and another object.
EQ statement option 3: coronoid process; Q: directed towards (posterior) [ this would be a pre-composed class called directed posterior’. It is relational but second entity (body axis) is implicit.
2. Is ‘orientation’ even correct for above example? [We decided that no, it isn’t] It is_a relational spatial quality and requires two entities.
Question: Is a second entity (such as ‘body’) implied?
–E: coronoid process + Q: orientation^has_quality(posterior) RE: body
Another way to represent this using post-composition of ‘position’ (which is monadic)?
Action item: (Chris and George) We will add the necessary pre-composed classes to PATO, e.g. “directed posteriorly”. The definitions for these will be derived from the spatial ontology.
Other stuff: To fix in PATO: ‘anterior’ should not be a synonym of ‘anterior to’ as it is now. ‘Anterior’ is defining a region of the body vs. anterior to, which is relational. Also see posterior and posterior to and probably others.
Example of anterior used to describe a region: Buckup 1998, #75: olfactory bulbs positioned anteriorly”: EQ: E: olfactory bulb; Q: located in; RE: anterior region part_of body
To fix? why is anterior_to a syn of ventral_to (mark as human torso as others?)
Result: George fixed this. (i.e. he removed the synonym ‘anterior’)
Spatial ontology (BSPO) vs. PATO[a]: Spatial terms are inconsistently used within PATO: E.g. ‘curved anteriorly’ (PATO: 0001466) but not ‘arched anteriorly’ (or arched in any other direction). Also, no ‘curved antero-ventrally’ or any number of other spatial terms. Shouldn’t these be represented with PATO + Spatial in quality post-compositions or have all possibilities pre-composed? e.g., ‘curved’ (PATO:0000406) + anterior (no such term, however, in BSPO).]
Action item: (Chris and George) If these pre-composed terms are created (TBD) as classes in PATO then the logical definitions must be generated (using spatial.obo + pato.obo) -- also, review existing pre-composed spatial PATO terms and provide logical definitions if needed. Diagrams, drawings, examples, etc. could be added to help illustrate definitions.
Visual perspective: Axes differ across taxa, so and thus ‘dorsal’ views, e.g., are not comparable. Handled already for ‘rostral/anterior -- vertebrates/humans’, but not generally. E.g., bone has a rectangular shape from a dorsal view. simple fix could be dorsal surface of bone has rectangular shape -- but this won’t always suffice. e.g. bone A overlaps bone B from a lateral view. or bilaterally paired processes form a fan shape when viewed from dorsal perspective.
3. Paula: Size vs. shape :
General mismatch between natural language, the meaning of an author, and pato terms. Can a relationship be added between size and shape within PATO to handle this or is this an interface issue only? Size part_of shape? E.g., a relationship between short and shape, would return all the annotations.
Action items below important to all involved (ZFA, Phenoscape, HAO, plants):
Action Item: move “elongated” to make it a child of shape. [done]
Action Item: relabel ‘dwarf’ to ‘dwarf-like’ or dwarfed? (Dwarfed is common in plants) [done]
Action Item: Move “stubby” to make it a child of shape. [done]
Action Item: Review all children of size and see which ones need to be moved to be children of shape (e.g. narrowness is a shape, as is thin). [we decided not to move these but add comments to the relevant qualities] DON’T SEE COMMENTS - WOULD BE HELPFUL TO KNOW WHY NOT MOVE
Action Item: Check whether 3-D extent (parent) should be collapsed with volume (child)
Action Item: Check whether 2-D extent (parent) should be collapsed with area (child)
Action item: obsolete 1485; create two new terms: ‘condensed’ and ‘compressed’ with two new IDs. “condensed” (child of structure) more closely packed together; pressed tightly together;
Action item: Move ‘flattened’ and all of its children from Curvature to Shape. Add ‘compressed’ as syn for flattened. Get curvature out of defs of children of ‘flattened’. [done]
Action item: change ‘morphology’ def to read ‘shape or size or structure’ [done]
Size def: A morphology quality inhering in a bearer by virtue of the bearer's physical magnitude
Shape def: A morphological quality inhering in a bearer by virtue of the bearer's ratios of distances between its features (points, edges, surfaces and also holes etc).
Free text marked up to show PATO:shape and PATO:size terms.
Problem is that a user searching for shape will not return these annotations:
1.Basibranchial 3, shape: short and wide, almost square [Armbruster 2004, #4]
2.Shape of lateral ethmoid: short [Britto 2003]
3.Hyomandibular (otic region joint) shape: short, cranial facet less than twice depth [Lundberg 1992, #50]
4.Dermethmoid shape: short and wide [Smith 1992, #18]
5.Shape of ischiac process: small or posteriorly elongate [Britto 2003, #70]
6.Interopercle shape: elongated [Lundberg 1992, #77]
Action item: elongate moved to shape [done]
7.Urohyal shape: reduced to small bony nodule [Friel 1994, #38]
8.Shape of cleithrum: reduced: lateral and ventral laminae are reduced [Sawada 1982, #32]
9.Shape of basibranchial rays: anterior rays narrow [Albert 2001, #102]
10.Anterohyal shape: greatest width less than half of length
11.Shape of ceratobranchial 5: thin, uniform width [Armbruster 2004, #10]
12.Interopercle shape: expanded posteroventrally [Lundberg 1992, #77]
13.Palatine shape: large, thickened, massive [Santini & Tyler 2003, #42] -
14.Shape of branchiostegal rays: anterior 1-2 rays broad [Albert 2001, #145]-
15.Shape of anterior ceratohyal: anteriorly narrow [de Pinna et al. 2007 #84]-
Action item (George): Comments for size related qualities [done]WE DON’T SEE ANY COMMENTS
See powerpoint for examples. For example: “slender rib” (is_a shape) vs. “short, stout rib” (is_a size); or “Relative length of anterior process of vomer : 0) slightly broader than long, or as long as broad; 1) longer than broad” <- Example of possible EQ statements for these?
-- (also brings up issue of logical independence vs. biological independence)
--authors use language for size and shape that is inconsistent with pato. slender is a shape but elongate is a size... def for elongate: length is higher than width - ratio - a shape. Also ‘stubby’. [these are wrong, see above action items]
‘Compressed’ (synonym of ‘condensed’ PATO:0001485) is_a structure - def: A structural quality inhering in a bearer by virtue of the bearer's being thicker or more closely packed together; pressed tightly together. ‘Compressed’ used to convey ‘shape’ difference more so than diffs in ‘structure’. Word used in different senses by different authors/communities. We would expect compressed to come up in a search for shape.
Cavender #24: Basiocciptal process, compressed. Currently Q is ‘size’.
Friel 1994, #27: Shape of posterior end of palatine: compressed laterally with simple rounded margin. Here the author uses ‘compressed’ to describe the shape of a structure.
Britto 2003, #42: Shape of palatine: compact. ‘compact’ is a synonym of condensed, which is_a structure. Again, the author uses a child of ‘structure’ to describe shape.
4. Paula. Surface feature shape (PATO:0001925) vs. texture (PATO:0000150)- what is the difference?
These terms have similarities; hard to say how they are different.[b] Texture is a surface feature shape? -- is anything with a texture not smooth? Scute, smooth [child of texture] vs. scute, ornamented [ where the author means at least not smooth. Could be type of ‘surface feature shape’ or type of ‘texture’. Why is ‘surface shape’ a sib of terms like split and striated? doesn’t seem logical.
More on ‘surface shape’: better as post-comp?: ‘Surface part_of Entity: flat’ vs. current Entity: flat is_a ‘surface shape’)
EXISTING PATO:
+++++is_a PATO:0000052 ! shape
++++++is_a PATO:0002004 ! surface shape
+++++++is_a PATO:0001925 ! surface feature shape ***
++++++++is_a PATO:0001359 ! rugose
++++++++is_a PATO:0001365 ! spiny
++++++++is_a PATO:0001367 ! lobate
++++++++is_a PATO:0001907 ! botryoidal
++++++++is_a PATO:0001910 ! folded
++++++++is_a PATO:0001976 ! serration
+++++++++is_a PATO:0001206 ! serrated
+++++++++is_a PATO:0001368 ! erose
+++++++++is_a PATO:0001889 ! scalloped
+++++++++is_a PATO:0001975 ! unserrated
++++is_a PATO:0000051 ! morphology
+++++is_a PATO:0000150 ! texture ***
++++++is_a PATO:0000066 ! pilosity
++++++is_a PATO:0000700 ! rough
++++++is_a PATO:0000701 ! smooth
++++++is_a PATO:0001804 ! scaly
++++++is_a PATO:0001805 ! flaky
++++++is_a PATO:0001849 ! abrased
++++++is_a PATO:0001928 ! blistered
++++++is_a PATO:0002010 ! looseness
++++++is_a PATO:0002012 ! coating
++++++is_a PATO:0002255 ! grooved
Chris: 2-d vs. 3-d shape justification: cells round (2-d) vs. spherical (3-d); cross-sections of 3-d are 2-d, anatomical entities;
TENTATIVE EXPLORATION: Needs to be use-case driven -- what do we want to do with PATO? HAO: want to differentiate between taxa using qualities. Others: to cluster similar qualities that may have a genetic basis (e.g., invaginations and evaginations).
shape
+ 2-D (area) shape (aka cross-section, aka 2-d projection)
+ volumetric shape (=3d?)
++3-d convex
++3-d concave
+“linear” shape (=edge)
++ curves has an inflection between two bounding points
++ angular - has a particular type of inflection between two bounding points
shape pattern (repeated lines, areas, volumes; patterns have repeated shapes): grooved, reticulate, polkadotted;
--
- perspectives
- “magnitude”
Action items (George) (high priority):[c]
-add ‘cross-sectional’ and ‘2-d projection’ as syn for 2-d shape [done]
-add ‘3-d shape’ as a child of shape [done]
‘3-d shape’ will have child ‘3-d convex’ [done]
‘3-d shape’ will have child ‘3-d concave’. [done]
-delete ‘surface shape’ [done]
-keep ‘surface feature shape’ as child of shape: has some repeated feature of the surface; [done]
-add ‘has_repeated_part’ relation
-add ‘edge shape’ as a child of shape (e.g., serrated) [done]
-move ‘concave’ and ‘convex’ to more abstract, caro-like ontology level
Possible action items:
- invaginations vs. evaginations (protrusions) - understood developmental bases
- devise use cases (dream application: drag and drop (i.e., postcompose) concepts from PATO and annotated images = draw reticulate, evagination, and rounded, then overlay on image with scale-bar, etc.; we want to do this to delimit species that have slightly different patterns of surface sculpture)
Action Item:
Hold shape jamboree (possibly in association with RCN) to develop shape. Preparation requires generation of use cases: i.e. Matt (hymenoptera), Sandra (human bones), Wasila (fish bones). Also need information from developmental biologists (plants, insects, vertebrates) so that we can implement an upper level shape hierarchy that may allow us to cluster variation (mutant, species) according to qualities with similar known genetic bases.
Chris and George to carry out as much obvious clean up as possible in PATO prior to the meeting. RCN group responsible for identifying outside expertise.
HAO -> Our sculpture terms are similar to issues in 4: We have many terms for types of cuticular sculpture (e.g. reticulate, areolate, aciculate, colliculate) that have been previously defined in a shallow hierarchy of sorts. Our thinking of how to ontologize these concepts would benefit from an ontology of patterns- this may come largely from an outsiders perspective/work.
Does anyone know of this effort http://www.aimatshape.net/ ?
Need more hierarchy in ‘shape’
Concave 3-D shape and convex 3-D shape: Not sure how useful these parent terms are because it would seem as though almost any shape can have a convex or concave component at some level. For example, ‘lobed’ or ‘trilobed’ (is_a shape) could be children of convex 3-D shape. Also, ‘tripartite”, with children ‘T-shaped’ and ‘Y-shaped’, is a child of concave 3-D shape but it doesn’t seem like the concavity that T-shaped or Y-shaped might exhibit is an important way to classify these terms.[d]
Action item (George): Rework PATO “spatial pattern” hierarchy to reference other terms which the pattern is in respect to (structures, colors, etc?). [color patter moved to spatial pattern - we might want to consider bringing back the obsoleted term temporal pattern (sleep pattern)
5. Range values:
Andy: (originally from Hong Cui) We found it difficult to annotate a range value such as [stems being] "erect to prostrating", when an ontology defines erect is-a A, while prostrating is-a B. We can not logically annotate that statement using the ontology (just like we don't say "apples are green to big", a computer wouldn't know how an A can sit on the same spectrum with a B).Harder but even more helpful, if PATO can incorporate concept definitions to allow reasoning on the intermediate concepts in a range. For example, what concepts sit between erect to prostrating.”
Example from Hymenoptera:
head yellow to orange [slides] [= http://bit.ly/bRM4DI]
(i.e., some individuals of this species have more red than others, but all have yellow)
Recommendation from HPA: make discrete
Implement constraints via ... following up the path of relationships (up one attribute)...
Two issues:
a. How to represent range? Do it outside of PATO... representing something subtle -- an individual or a population.
Individual: The face of the wasp (a single one) grades from yellow to brown. New relationship needed?
Generally important (if support available, zfin and rest of us would use):
Action Item: (Chris and Jim, Phenex) define a relationship “ranges_from” which, in usage when annotating, specifies a continuum of values of the two indicated qualities.
This can be used in conjunction with color patterns (e.g., variegated) and other scalar properties (angles). But recumbent to erect or across shapes, very difficult. Add rules and formalisms so that when people ask ‘give me X with some part that is red’, it is returned.
Begin with scalar properties!
Populations: The face of the wasps range from yellow to brown. Technically the annotation would be the same as above (individual), but curator would indicate (outside of pato) what the range annotation is applied to (population, individual, etc.).
Action Item: (George) Add a grouping class for erect and prostrate (verticality?) under the superclass position. [but see item #14] [we need to alter the def but let’s look at it when we revisit position]
b. How to constrain the range to forbid nonsensical ranges? Want to know ‘can I make a range between red and green?’ Algorithmic way to implement such constraint? Use slim - values have to be inherited from same attribute. Go up to the first attribute that you hit, and if the same, can make a range.
6. Composite terms - requests and policies (slides)
Katja: “At the present time we are only collecting non-composite terms or one-word terms (rounded, translucent etc). We understand that its much more complex than this where adding a adverb to another term creates an entirely different meaning than the two terms alone. What are the guidelines regarding what kinds of new terms you will accept? I noticed in the request submissions there are some composite terms (ie curved medial, delayed closure).” Examples of composite terms we need: loosely attached, tightly attached
HAO: How do we best express modifiers, are these relationships between two PATO candidates? E.g. “narrowly something rounded”?
Paula/Jeff: 1. More on complex modifiers: examples of this come up very frequently in relation to size e.g. small, medium, large; or even something like--tiny, very small, moderately small, medium, moderately big, huge, etc. It’s very unclear how to capture this information given only increased size quality and decreased size quality (related to item #1).
=> also relate spatially. ‘curved anteriorly’ [e.g. Friel, 1994, #88: Antero-lateral margin of coracoid deeply concave, state 0 (vs. moderately concave, state 1) [currently eq’d both to ‘concave’.
Resolution: In general, precomposition is OK. Avoid adverbs. All terms need definitions.
Guidelines: for submission of terms to PATO: Be sensible! (submit precomposed terms as necessary for use by the community) E.g., the burden of definition could be a limiting factor. In general, do not add adverbs (e.g., gently rounded). Provide context through examples and counterexamples of use, which go into the comments, if possible.
Action item (George, Chris, HAO=research): Demonstrate how this could be done with a color model. Create an RGB for color representation - check for standards. Colors will be defined by RGB and wavelengths. Automatic classification based on values. We can provide color swatches, etc. to illustrate definitions. Identify the minimum information required for submitting a term to PATO, and publish this guideline somewhere.
Info on RGB and wavelength : http://www.efg2.com/Lab/ScienceAndEngineering/Spectra.htm
7. Term additions and definitions to PATO:
Action item: (George) Write guidelines for minimum information needed to submit new terms to PATO. Guidelines would be posted on PATO main page.
Action item: (Curators) Add definitions where possible for new terms.
Action item: Update applications to output to comments: 1) notes on context; 2) notes on definitions themselves.
Action item: HAO will prioritize (heatmap, workshops?) their 900+ terms for inclusion in PATO. Provide sentence of context.
Matt: Can we add all adjectives in Websters or wordnet?[e] (Gets to the minimum “evidence” needed for inclusion in PATO- is this evidence published, from an “expert” etc.)
HAO: How best to add a large number of new terms (700) with definitions and context?
Need a dictionary definition, but it is not sufficient. Genus differentia definitions needed for PATO? Very few genus-differentia definitions in PATO currently.
HAO: What exactly is a sentence of context? Where does this get persisted in, say, an OBO file? We should supply context (would go in Comment field).
HAO: Do we need to capture instances of the word usage as well as the definition when proposing a new term to PATO? (‘yes’ from S. Lewis).
Phenoscape wonders whether examples of usage could be added to comments? E.g. in PATO, ‘expanded’ is a syn of ‘distended’. But distended is not a good term for author usage. Distended references flexible stuff (abdomen) vs. bone. Expanded in our context is a better syn of ‘increased area’. Curator knowing prior usage would be helpful.
Hong Cui: I am also curious about the possibility of adding computer readable definitions to PATO (i.e. by using necessary and/or sufficient conditions, for example a necessary condition for erect is straight). I'd like to hear about how valuable PATO group thinks those definitions are and if they have plan to add those. <- The underlying issue is should all definitinions be in the form of GDif statements.
8. Domain specificity, PATO, and slims: “The CARO of PATO”
Andy: 2. If the criteria for addition to PATO is based on utility (for entity/quality statements etc) should we create domain specific PATO(s)? Answer no (Lewis), but axes, e.g., differ between organisms. Do we need a Phenotype Reference Ontology (PRO)? It’s reasonable to assume we need to split anatomical terms into multiple taxon-specific ontologies, but why not also phenotypes?
example1: osseus, ossified in botany = bony, becoming bony
example2: irregular in botany = bilaterally symmetrical but flower parts of different sizes/shapes
[slides] [= http://bit.ly/bRM4DI]
Q: How to do community specific synonyms? OBO - syn scope and syn types, so we would have syn type ‘community specific syn’ with multiple definitions. Their software would need to take advantage of that. E.g. peach pit example, ‘stony’ would be the preferred term, and ‘osseus’ would be plant community specific term (with same ID as ‘stony’). The software can be programmed so that community specific synonyms are applied correctly annotation of particular taxonomic groups.
Action item (required for anyone submitting such terms): Add “community specific synonym (taxon specific?)” functionality to applications AND add taxon-specific labels to PATO.
9. Do we need best practices for qualities that might better be represented in an anatomy ontology?
For example, how to represent “multicuspidate tooth”? Create new term in the anatomy ontology? Create a cross product term with “tooth” from anatomy ontology and “multicuspidate” from PATO? PATO has multicuspidate - but doesn’t this belong in an anatomy ontology?
Resolution: No - if we want to add ‘multicuspidate tooth’ to TAO, create Xproduct with PATO.
Action item (George): move tricuspidate to be a child of multicuspidate [done]
HAO has ‘adreniform’ for adrenid bee habitus (belongs in an insect quality ontology? - that descends from a branch of pato?). HAO-specific action item: eliminate andreniform, etc.
Resolution: Yes - add it...
Also, PATO:cartilaginous and PATO: ossified -- should these be cross-referenced to ‘cartilage’ and ‘bone’ terms in an anatomy ontology?
10. Disjointness and PATO (representation of negation and reasoning using disjoint_from):
Jim: Phenoscape + Chris: Many statements are negations, e.g., ‘not round’;
PATO needs to identify that round disjoint_from square to answer query ‘find all phenotypes with entity A and not PATO round’. Other e.g., ‘Bone A, not big’, curated how? ‘disjoint from’ is not in relations list. 2. Antonyms also seem to be an issue of disjointness (e.g., long vs. short).
Another query: Here’s a phenotype for this important gene -- what species phenotypes can I rule out (i.e., what phenotypes are not the same)?
Sawada #45: Enlarged pleural rib, absent, current EQ: ‘Rib, Size’, but should be negation: ‘Rib, not enlarged’.
Jim: for size, use binning system that we worked out yesterday (add # here)
Jim has implemented something in Phenex (describe yet)
Chris: need disjointness axioms. Hard for shape (not possible now).
Action items (High priority)[f]
Action 1: (George):
-For ‘un’ terms and ‘non’ terms, like ‘unserrated’, where possible add ‘positive’ term (e.g., uninterrupted is continuous); add ‘not [term] syn; [this is almost always the case - uninterrupted is a synonym of continuous in PATO which is the preferred name - we need terms such as non-ictal this is the common term name in biology)
-make ‘un’ terms consistently un- or un (dashed or not) [done -this was more for non rather than un - two cases fixed)
Action 2: (Jim): Jim will implement negation in Phenex using ‘complement_of’. Jim will add view of ‘disjoint’ to Phenex interface (so curator knows whether to request that a term be disjoint)
Action 3: (George): check terms for missing disjoint_from axioms [ongoing task - there a lot of disjoint axioms in PATO - when we look shape more carefully we could add more of these_
Paula asked whether there was any way to find out whether kbs are using particular terms - there is not currently the infrastructure to do this.
Un- from PATO candidates in HAO
unarmed
unbranching
undefined
undeveloped
unenlarged
unequal
uneven
unmodified
unnotched
unsclerotized
unsculptured
untinted
11. Need to say that two entities are the same vs. different:
Matt: For e.g. we want to express that two parts are concolourous (share the same color) or they don’t- how do we indicate the latter using PATO? Aside- both “color” and “colour” are used in PATO, just like here. e.g. ‘head and mesosoma concolorous’. Relational quality? A concolorous with B?
Phenoscape examples for many types of qualities:
1. Sidlauskas and Vari 2008, #122: Presence or absence of small dark spots positioned at centers of scales; centers of scales same color as, or lighter than, scale margins
2. Friel 1994, #118: Jaw occlusion: 0, upper and lower jaws project equally and mouth is terminal; 1, lower jaw is subequal to upper jaw producing a distinct overbite
3. Toledo-Piza 2007, #80: Type of contact of ventral portion of mesocoracoid with coracoid and scapula: 0, with greater contribution of coracoid, with posterior extension of ventral portion of mesocoracoid; 1, with approximately equal contributions of both scapula and coracoid
4. de Pinna 1993, #199: Neural spine of the second free vertebra: 0, similar in size, shape and orientation to other anterior free vertebrae; 1, in plotosids only, the neural spine of the second free vertebra is markedly enlarged and strengthened relative to its neighbors, and also contrasts with them in its directly vertical orientation (vs. posterodorsally as in the primitive condition) and very deep distal bifurcation
General upper-level relational quality “has_same_value”? post-compose with “color”, etc? Example with made-up relation: A has_same_value^relative_to(color) towards B
Phenoscape current practice:
Example “frontal bone larger than parietal bone”
–E: frontal bone + Q:size^increased_in_magnitude_relative_to(parietal bone). [needs to be ‘size inhering in parietal bone’]
Terminology:
E : the bearer of the quality
Q : the quality / attribute
Dependent Entity : only used for what is currently called the “relational quality” branch of PATO (though we may want to change the name here)
* other choices: contingent entity or co-requisite entity
ComparativeReferent : only used when value comparisons are being made
Option 1 (pseudo interface, don’t have to do any cross-product)
PhenoSyntax:
E= metasoma Q=color ComparativeReferent= mesosoma ComparisonReferentRelation= equivalent
Translates to ==>
(color and inheres_in some metasoma) equivalent_in_value_relative_to (color and inheres_in some mesosoma) <- how persisted in OBD
non-Option 2 (shortcut, how it is done thus far, semantically equivalent to Option 1, the difference is PhenoSyntax:
that below needs ‘color and inheres_in metasoma’)
E: metasoma; Q: color^equal_in_value_relative to(mesosoma)
Action item (George, high priority):[g] add more relations to PATO to generalize existing scalar quality comparisons to any kind of value comparison
-check with NIF folks first
equivalent_in_value_relative_to << CHANGE NAME OF EXISTING REL
different_in_value_relative_to << CHANGE NAME OF EXISTING REL
increased_in_value_relative_to << CHANGE NAME OF EXISTING REL
decreased_in_value_relative_to << CHANGE NAME OF EXISTING REL
Action item: specify logic such that Option 2 is equivalent to Option 1
(color and equal_in_value_relative_to some mesosoma) inheres_in some metasoma)
PATO hierarchy:
color
equivalent color relative to
different color relative to
equivalent color relative to = color AND equivalent_in_value_relative_to SOME color
different color relative to = color AND different_in_value_relative_to SOME color
Action item (George): add these pre-composed classes to PATO, with logical definitions
Action item (George): add pipeline to create slims that exclude pre-composed relative to comparative classes
***
Polkadots example: ‘number of dots less dense on head than torso’. Interested here in representing the spacing between the dots.
Suggestion to use an AO class e.g., patch and a “spatial density” quality (not to be confused with the existing “mass density” class in PATO).
Action item (George): add and define “spatial density”: def: number of objects per unit area or volume.
12. No definition/obsolete high level term ‘qualitative’:
Matt: HAO: 3 children of the root, one of which is “sort of” obsolete but not dealt with, “qualitative”. This term has many children.
PATO aren’t dealing with these very high level issues I see little hope for practical use of PATO by those of us with less experience.
rationale: pato uses bufo as an upper level ontology...everything real...but people like to use ‘absent’ which is hard to handle.
Action item (George): define ‘conspicuous’ and move to quality of object branch
Action item (George): review existing terms under qualitative
Action item (George; high priority): document the qualitative branch in the comments field. e.g. “this is intended for use as shortcuts”, possibly change the preferred term label to something more indicative of what child terms might be found here.
13. Confusion between relational qualities and relationships:
Jim: How should similar qualities and relationships be used? Do they need to cross-referenced to one another? For example, a character might be “ structure A located in structure B”. In EQ syntax, this is E:structure A, Q: located in, E2: structure B. But there is a “located_in” relationship as well. PATO as ontology of owl:properties?
There is redundancy between relations and pato qualities - could curate a phenotype w/out a quality (e.g., ‘E located in E’ -need no Q); do we need to keep it?
Action Item: Chris: look towards moving away from using these relational qualities once we implement taxon-specific relations for anatomical entities. Right now the relational quality is the only way to indicate a variable phenotype.
14. Reorganize PATO to move relational qualities in with their related monadic qualities. Was: More human-readable names for various terms?
E.g. “relational spatial quality” can seem jarring when used as a character type in the Phenoscape KB web interface.
relational spatial quality, relational structural quality, relational shape quality, etc.
Action item (George); highest priority:[h]
* Eliminate or merge grouping terms “relational X” into monadic branch.
* This may involve merging currently disjoint parts of the hierarchy, e.g. position and relative position.
* Provide a different means of specifying that a quality depends on an additional entity.
Action item from #2: (Chris) Look at what is under position. Might remove position after considering subclasses. [subsumed under # 14]
Request: (Suzi) eliminate the string “quality” from as many terms as possible.
Paula & Wasila suggest ‘orientation’ in place of ‘angular placement’.
Action item: change primary term from angular placement to ‘orientation’ [done]
Action item: renamed “oriented” (currently under regular spatial pattern) and define appropriately (e.g. a pattern where all the repeated elements are oriented in the same direction)[done]
Action item: move edibility to disposition [done] [PATO is a mxture of dispositions and qualities which are tagged via the disposition slim] Maybe all of these need to be moved.
Action item: move mobility to disposition [done]
Action item (for sometime...): review Physical qualities
26 September 9:00 am - 6:00 pm
15. Paula. Color /pigmentation:
See slides for examples. Color and pigmentation are often used synonymously in the literature, but they have different meaning. Also, “color” and “degree of pigmentation” have different parents in PATO, but are related to each other (pigments produce color...). Use degree of pigmentation terms with entity (as zfin), e.g. ‘anal fin pigmented’ would be E: anal fin, Q: pigmented)
How to classify patterns of pigmentation/color (e.g., bars, stripes, dots)? -- see #1 below.
Action items (George):[i]
(1) move color pattern to be a child of spatial pattern. [done]
Request pre-composed ‘banded color pattern’, and analogously may (on demand) pre-composed ‘banded xxx pattern’ (where xxx is the physical thing establishing the pattern);
add ‘banded’, ‘striped’, and ‘punctate’.
(2) move crayola terms currently under “color” to be children of hue.
(3) hue, saturation, and brightness would be direct children of “color quality”
(4) discolored, colored, colorless, variable color (move to spatial pattern), [done] variability of color (and its two children, which deal with temporal variability) should be placed elsewhere in the ontology
(5) variability hierarchy reflects temporal variability, and it should be indicated in the definition (for variability of color)
6. TAO needs to obsolete ‘pigmentation’
7. Add “correlates with”(?) between “degree of pigmentation” and “color quality”.
16. Other important questions:
--pato subdivides population vs individual qualities.
Action item (George): viability and its children should not be subclasses within “population quality”. Move to ‘organismal quality’ [done]
not really, but can put a xref to a thread (not done in pato yet, but examples from GO), e.g.:
http://amigo.geneontology.org/cgi-bin/amigo/term-details.cgi?term=GO:0001909
links to http://www.geneontology.org/cgi-bin/references.cgi#GO_REF:0000022
Action item: clean up this doc and publish on google (Wasila); add uri to comments block to each term that is changed (added, obsoleted) and term relationship changes (so add uri to each term involved in change).
Action item (George): add ‘centered’ to PATO (not bspo because that has relations) with definition: positional quality...with bearer located in a position equidistant from edges. And consider this with respect to the existing term “medial” (“...bearer's being located toward the middle relative to another entity”). [done]
e.g., pigment patch is centered in sp A vs. not centered in sp B
Phenoscape annotations with ‘center’ in reference to scale - just add to TAO ‘scale center’ (there might be a specialized term for this - check)
This is because early on there were lots of changes in PATO. Not to add retrospectively.
Action item:
-moving forward, need to give reasons for obsoleting.
II. EQ Representation:
17. Complex phenotypes: how far should we take EQ?
Topic for discussion - what does user expect to see? How does this guide best practices in terms of tools to generate data? For example: Horizontal expansion of anterior ceratohyal. 0) anterior ceratohyal, mainly vertical laminar-like element 1) extended medially to form thin horizontal laminar-like expansion. Horizontal expansion broader than vertical element" or “posterior margin of fourth infraorbital distinctly angled anterodorsally at least on posterodorsal portion of bone, with width of dorsal portion of bone distinctly narrower than ventral margin”
Discussion: Depends on use. For species descriptions need to be very granular. discussed trade-offs between complexity of annotations (postcomps) vs. simple but highly granular annotation (e.g. very specific and highly granular term).
18. Ternary relations/Phenotypes involving 3+ entities? For example: Pelvic fin position: midway between pectoral fin-base and anus
Hong Cui: How do you deal with ternary relations in EQ? For example: "Organ A distant from Organ B by an Organ C's length"?
Nothing required, working : EQ solution for above: distance^inheres_in(organA)^towards(organB)^similar_in_magnitude_relative_to(length^inheres_in(organC))
Phenoscape currently annotates these to a coarse level. In the example above, the EQ would be E: organ A, Q:distance, E2: Organ B. In other cases, we record two EQs: for example, Organ A located midway between Organ B and Organ C:
E: Organ A, Q: relational spatial quality: E2 Organ B
E: Organ A, Q: relational spatial quality: E2 Organ C
examples
Ishii & Imamura 2007 19 First basibranchial present, connected with second basibranchial through cartilage basibranchial 1 bone relational structural quality basibranchial 2 bone relational structural quality
19. Counts and ratios:
19a. How to represent standard anatomical measurements, typically recorded as a ratio?: Phenoscape: For example, Head length >19% Standard length. These are currently annotated at a coarse level (E: head, Q:length). Question is whether the anatomy ontology should also contain terms for standard measurements in ichthyology? I’m sure this is an issue in other fields as well.
HAO: In our field “ratio” and ratios are very frequently used. These aren’t ratios, i.e. for 5:4:4 the ratio is 10:8:8. We’ll need to best figure out how to construct ratios within markup. Similarly- we often use the concept of a “formula”, e.g. antennal formula 15:14, tarsal formula 5:4:4 (<= i.e., fore tarsus with 5 tarsomeres, mid tarsus with 4 tarsomeres, hind tarsus with 4 tarsomeres); labial formula 3:2. These are just ratios, sometimes complex. <- Seems like an application level, “could be formalized”.
Discussion: Chris: plant ontology folks interested; George: ratios are not qualities, but rather mathematical expressions (Chris: but so are squares...). George: does not recommend pre-composing- in general we need to have better representations, this is not only for phenotypes.
Example of how to do this currently:
Free text: Head length 55% of Head width.
EQ Syntax:
E=head Q=ratio^has_dividend_quality(length)^has_divisor_quality(width) V=55 %
from plant TO:
[Term]
id: TO:0000298 ! chlorophyll ratio
intersection_of: PATO:0001470 ! ratio
intersection_of: has_dividend_entity CHEBI:18230 ! chlorophyll a
intersection_of: has_divisor_entity CHEBI:27888 ! chlorophyll b
intersection_of: has_ratio_quality PATO:0000070 ! presence or absence in organism
Equivalent, in OWL manchester syntax:
‘chlorophyll ratio’ = ratio and has_dividend_entity some ‘chlorophyll a’ and has_divisor_entity some ‘chlorophyll b’ and has_ratio_quality some ...
Action item (low priority): George will develop and send out mathematical solution to this problem.
19b. How do you record for example counts, values etc.?
Number of antennomeres 4-6.
Length of head 1.22 mm.
Body length over 2 mm.
Average body length 1.2 mm +/- 0.234mm.
Units - There is a units ontology, it maps to the qualities in PATO. Ratio is a dimensional unit. Use the has_value, take the quality, add the value , give the number, select from PATO size and length. So the Units ontology will flag mistakes, like “length 20 grams”.
20. Inheres_in:
HAO: As I understand it the Phenoscape folks are using inheres_in to link PATO to their ontologies. Is this a “best practice”? What are the other possible relationships/properties?
Phenoscape response: We use inheres_in to express the a phenotype as quality that inheres_in anatomical structure. Do you have a quality that does not inhere in an entity to instantiate a phenotype? - hlapp <- Maybe - a caddisfly stone/twig case (= part of extended phenotype, i.e. it’s a behavior?)
Where do we put bee’s wax, mutualistic fungi, honey (=vomit), silk, wasp nest (saliva and mud), ant mounds, other stuff? Maybe Envo as an example.
Action (HAO): Explore bee’s wax, wasp paper etc. as refernce in HAO but stored in ENVO.
21. Should there be a general purpose code library for constructing EQ statements, like rubyEQ, Bio::EQ or jEQ? What about development of some rules for persisting EQ statements (“EsQl”). -- Matt
- Q: Should PATO definitions strive to be in GD format? A: yes, ideally, but they look strange to user communities.
EQ syntax is a template -> generic level OWL APIs. Template level. Provides the common logic. Maybe a webservice that allows you to specify some values-> are these values valid? provide a definition, use these in a pre-composed phenotype term.
22. Sets of entities:
From October 2008 PATO meeting notes:
Where one member is compared to a set or series. For example, “Vertebra 1 is largest of all vertebra”. Chris says we could make a new relation that would have complex 1st order logic but could be simple to use. For example, you could say:
vertebra1 increased_in_magnitude_relative_to_all_members_of_set_of_X_in_same_organism
owl manchester syntax: vertebra 1 is smaller than (i.e. size decreased in magnitude relative to and posterior to) no vertebra
Refer to the nth vertebra with n-1.
Two solutions, ennumerate all units (vertebrae, flagellomeres), or in the case where you don’t know you can use a set approach.
Action (Jim, Chris): implementation to consider
Question: Similarly, how to represent the first/last member of a series?
A: add ‘posteriormost vertebra’ to TAO; posteriormost vert is anterior to no vertebra is smaller than no vertebra;
--’sequentially smaller’ part of the statement would have to be enumerated (but not possible if you don’t know the number of vertebrae); if x is anterior to y, then x is smaller than y.
all vertebrae are smaller than and posterior to no vertebra;
Logically more complex logic to represent general case (this expression will slow down reasoning). Do we need to reason over these? Use case...
Example: Angular difference between the bases of the first and last branchiostegal rays
0, about 90 degrees; 1, oriented approximately at a 135 degree angle to one another [de Pinna 1996, #51]
? E: branchiostegal ray first_in_series_of(brachiostegal ray series)
[a]no xrefs to BSPO in PATO? what is the history/current use of this ontology? - adeans
[b]single feature vs. repeating elements? - peter.midford
[c] High priority (#2)- pmabee
[d]Seems that both of these are the same unless we have a well resolved orientation framework - matt (see 6)
[e]Wouldn’t that clutter up PATO with a lot of questionably relevant terms (e.g., ‘cheerful’). Would need a lot of curation effort to filter irrelevant or metaphorical usages. - peter.midford
- pmabee
[f] High priority #3- pmabee
[g] High priority #2- pmabee
[h]Highest priority (#1) - pmabee
[i]High priority- pmabee