search

obophenotype / upheno

The Unified Phenotype Ontology (uPheno) integrates multiple phenotype ontologies into a unified cross-species phenotype ontology.
https://obophenotype.github.io/upheno/
Creative Commons Zero v1.0 Universal
76 stars 17 forks source link

Representing complex phenotypes #590

Open matentzn opened 4 years ago

matentzn commented 4 years ago

There turns out to be quite a number of what I want to call "complex phenotypes". A complex phenotype is a phenotype whose definition involves more than one entity. So far the Phenotype Editor have identified the following major categories:

We need to collect all such cases here to develop a strategy on how to handle.

matentzn commented 4 years ago

Excerpt from emails with @mah11:

I've looked for "causal" phenotypes in FYPO, and I have not found any
where I've tried to do logical definitions (I wouldn't have known how).
But I did find a few things that might be good to look at, including
these examples:

id: FYPO:0000913
name: abnormal sporulation resulting in formation of ascus containing
non-uniform spores
def: "A sporulation phenotype that results in the formation of an ascus
that contains spores of non-uniform size and DNA content." [PMID:20298435,
PomBase:mah, PomBase:vw]

"spores are non-uniform because sporulation was abnormal"

id: FYPO:0001861
name: increased minichromosome loss upon segregation during vegetative growth
def: "A cell phenotype in which minichromosomes are lost due to abnormal
mitotic sister chromatid segregation at a higher frequency than normal."
[PomBase:mah]

"chromosomes are lost because sister chromatid segregation went wrong"

... and there's this, and a couple more like it, where the causes are in a
comment, not even in the text definition, because there's more than one
possible:

id: FYPO:0003604
name: sister chromatid separation during meiosis I
def: "A cellular process phenotype in which sister chromatids become
physically detached from one another during the first meiotic nuclear
division, instead of remaining attached and segregating together. Sister
chromatids normally do not separate until the second meiotic nuclear
division." [PMID:10440376, PomBase:mah]
comment: Sister chromatids may separate in meiosis I due to problems with
either cohesion or kinetochore attachment to microtubules. Therefore
consider also annotating to 'abnormal attachment of spindle microtubules
to kinetochore involved in homologous chromosome segregation'
(FYPO:0000209), 'decreased meiotic sister chromatid cohesion'
(FYPO:0002093), or 'merotelic kinetochore attachment during meiosis I'
(FYPO:0005638).

This means "sister chromatids separate too early, because either sister
chromatid cohesion or centromere attachment to the kinetochore (or maybe
both?) was abnormal".
anna-anagnostop commented 4 years ago

Examples of "complex phenotypes" from the MP Ontology: @sbello

periodontal pocket MP:0030490 DEF: a pathologic deepening of the gingival sulcus as a result of apical migration of the junctional epithelium and the destruction of alveolar bone and periodontal ligament fiber bundles

hemolytic anemia MP:0001585 DEF: deficiency of red cells resulting from an increased rate of erythrocyte destruction

acidosis MP:0003031 DEF: a pathological state characterized by an increase in the hydrogen ion concentration in tissues and blood caused by an decrease in the concentration of alkaline compounds, or by a increase in the concentration of acidic compounds or carbon dioxide to the body fluids

(see also child terms such as respiratory acidosis MP:0012550 DEF: acidosis caused by retention of and increased blood carbon dioxide concentration, often caused by decreased ventilation (hypoventilation); results in decreased blood pH unless kidneys compensate by retaining bicarbonate)

emphysema MP:0001958 DEF: an abnormal condition of the lung characterized by permanent enlargement of airspaces distal to the terminal bronchioles, accompanied by destruction of their walls and without obvious fibrosis; results in decreased respiratory function including increased air retention and reduced exchange of gases (NOTE: current EQ uses MPATH term for emphysema)

aneurysm MP:0003279 DEF: a protruding sac formed by the dilation of the wall of an artery, a vein, or the heart resulting from a weakening of the vessel wall or heart muscle

peliosis MP:0013332 DEF: presence of multiple cyst-like blood-filled lakes or cavities that lack endothelial linings within parenchymatous organs, typically found in the liver, spleen, bone marrow and lymph nodes; also described in other organs such as lungs, kidneys, parathyroids and pancreas

cyanosis MP:0001575 DEF: a dark bluish or purple skin discoloration resulting from deficient oxygenation of the blood (current EQ: has part some (cyan and inheres in some zone of skin and has modifier some abnormal)

dosumis commented 4 years ago

For phenotypes that have a causal aspect to their definition, we need P -> P causal relationships we've already discussed adding to RO. Adding these is a priority https://github.com/oborel/obo-relations/issues/350

conjunctive phenotypes could make use of the has_part pattern, just adding further has_part clauses. I think this would give us the inference we need (e.g. ' elongated multinucleate aseptate vegetative cell phenotype' subClassOf 'elongated cell phenotype'.) I guess this means baking in the has_part pattern forever, but at this point I think I could live with this.

pnrobinson commented 4 years ago

I am not sure it is a great idea to add mechanisms to phenotypes. There are many that are relatively clear such as cyanosis. In many cases, the clinical appearance is so characteristic that one uses the mechanism as a shorthand to describe what one observed with a minimum of inference (e.g., with cyanosis). There will be lots of terms where there is a relatively fuzzy relationship with mechanism.

dosumis commented 4 years ago

On 14 Feb 2020, at 14:28, Peter Robinson notifications@github.com wrote:

I am not sure it is a great idea to add mechanisms to phenotypes. There are many that are relatively clear such as cyanosis. In many cases, the clinical appearance is so characteristic that one uses the mechanism as a shorthand to describe what one observed with a minimum of inference (e.g., with cyanosis). There will be lots of terms where there is a relatively fuzzy relationship with mechanism.

Where the relationship is fuzzy - we should not record it. The new relations will give us the option to record this relationship where it is safe to do so. Some of the examples above look safe enough (e.g. haemolytic anaemia).

chris-grove commented 4 years ago

Here are some potentially complex phenotypes from the C. elegans phenotype ontology. Please note that in some cases I think the terms should probably be edited to modify the label and/or definition:

WBPhenotype:0000162 pale larva Larva show deficiencies in chroma resulting in increased translucency.

WBPhenotype:0000215 no germ line Animals lack all germline progenitor cells, thereby resulting in the complete absence of a germline.

WBPhenotype:0000697 protruding vulva Animals undergo incomplete vulval morphogenesis, resulting in the formation of a single protrusion at the site of the vulva.

WBPhenotype:0000915 pale adult Adults show deficiencies in chroma resulting in increased translucency.

WBPhenotype:0001041 meiosis defective early emb Animals exhibit defects in the progression of meiosis which ultimately result in embryonic lethality.

WBPhenotype:0001384 fertility reduced Animals exhibit a reduction in the production of new individuals that contain some portion of their genetic material inherited from that organism as a result of defective gametes.

WBPhenotype:0001633 lethal gamete Animals produce ova/sperm that contain lethal factors as a result of chromosomal disjunction defects.

WBPhenotype:0001635 excess pharyngeal cells Animals contain an excess number of pharyngeal cells compared to control. In C. elegans, pharyngeal cells are derived from MS-blastomere. The overproduction of pharyngeal tissues is a result of other blastomeres adopting an MS-like fate.

WBPhenotype:0001105 P0 spindle position defective early emb Altered P0 spindle placement causes either a symmetric first division, a division in which P1 is larger than AB, or a division in which the asymmetry is exaggerated such that AB is much larger than normal.

WBPhenotype:0000660 social feeding increased Worms are found in clumps and/or crowded at the border of the bacterial lawn more frequently than control animals. This behavior is not due to mobility defects.

WBPhenotype:0001586 multiple anchor cells Animals contain more than one anchor cell. In C. elegans, presumptive ventral uterine cells undergo fate transformation and instead acquire the properties of an anchor cell (often due to defects in lateral signaling).

matentzn commented 4 years ago

This is a list of all 45 HP classes I could find that seem in some way complex (proxy: uses has part more than once, some false positives likely).. We need to categorise which categories they belong to:

anna-anagnostop commented 4 years ago

Additional possible MP Complex Union Phenotypes

sbello commented 4 years ago

Example from the MP mappings tab - hydronephrosis - in mp eq is has part some (dilated and (inheres in some renal pelvis) and (has modifier some abnormal)) But in HPO the same eq is used for the parent term Dilatation of the renal pelvis (HPO) Both ontologies specify in the text definition for hydronephrosis that dilation is in the renal pelvis and calices. In UBERON the calices are part of the renal pelvis but it is important to specify for hydronephrosis that both the pelvis and the calices are dilated.

sbello commented 4 years ago

Another example from MP mappings, HPO:0008873 shares an eq with MP:0000547 but the HPO eq is missing that the HPO term is short limbs with an average size trunk.

sbello commented 4 years ago

Since @chris-grove asked about a detached pattern I went looking in the MP and I think we need to add this to the complex list. Example term: detached Reissner membrane http://purl.obolibrary.org/obo/MP_0006023 current eq: has part some ( detached from and inheres in some vestibular membrane of cochlear duct and towards some scala vestibuli and has modifier some abnormal) eq specifies both what is detached and what that entity is detached from.

jseager7 commented 3 years ago

Late to the discussion, but PHIPO makes pretty extensive use of complex phenotypes. As part of our pattern mapping effort I've been trying to categorize the problematic phenotypes. Here's some examples; I can provide a comprehensive list if required:

Temporal phenotypes

These could probably be modelled with patterns that use terms from 'temporally related to' (RO:0002222), but that might depend on whether the subjects are always occurrents.

Causal phenotypes

These examples could probably be modelled using terms from 'causal relation between material entity and a process' (RO:0002595), but that may not cover all cases.

Spatial phenotypes

Some of these are probably definable using the '…in location' patterns, but that's usually complicated by the lack of variables (e.g. an ontology term for a host organism). Sometimes the location is further contextualized though, as with 'host lesion'.

Two phenotypes in one

I'd say that these phenotypes aren't even valid, since our annotation model supports (or should support) decomposing these phenotypes into individual phenotypes and adding them as separate annotations.

Doubly complex phenotypes

These are the worst-case scenarios, where a phenotype combines both causal information and spatial information. I can see potential cases where temporal information could be included as well. I'm really not sure if it's feasible to model these as precomposed patterns except on an ad-hoc basis, because the number of variables would cause a combinatorial explosion in the number of patterns.

matentzn commented 3 years ago

@jseager7 great to hear from you, and thank you very very much for this.. We wont look at this right away, but later this year, @rays22 will probably start looking into this. No easier answer, but your categorisation is super useful and helpful.

dosumis commented 3 years ago

@rays22 - we should discuss how to triage these. I'll work on a strategy with @matentzn for easier review.