In this guide, we will be concerned with the COVID-19 ontology, as it lives on http://www.aber-owl.net/ontology/COVID-19/
Refer to the installation guide.
In the COVID-19 ontology, many different aspects of COVID-19 are discussed. Of these, COVID Clinical Aspect (COVID:0000402). However, only one subclass of this is actually asserted in the ontology. We can see this if we run the following Komenti command:
$ komenti query -q "COVID clinical aspect" -o COVID-19However, if we look on the AberOWL website, there is a wealth of classery below this term. This is because they are inferred subclasses. They are inferred because terms we can see under COVID Clinical Aspect contain a subclassof axiom: isClinicalAspectFor some COVID-19. They are inferred as subclasses because COVID Clinical Aspect (COVID:0000402) contains an equivalency axiom for isClinicalAspectFor some COVID-19. Komenti can query these axioms and their subclasses directly:
$ komenti query -q "isClinicalAspectFor some COVID-19" -o COVID-19This will return a list of hundreds of terms across a wide range of specialities. Using the asserted and inferred structures of the ontologies that assert the original ontologies to further specify the set of terms that we're interested in:
$ komenti query -q komenti query -q "isClinicalAspectFor some COVID-19 and 'therapeutic procedure'" -o COVID-19 > labels.txtThe command collects different metadata associated with terms that are all clinical aspects of COVID-19, but are contained under different asserted substructures. Grouping them by their additional associated superclasses 'therapeutic procedure' we group all of these into one with the class description. Now, in labels.txt we have 6 labels for 5 different terms that describe therapeutic procedures for COVID-19. We can gain more labels for this by using Komenti's inter-ontology synonym expansion function (see here for more info):
$ komenti query -q komenti query -q "isClinicalAspectFor some COVID-19 and 'therapeutic procedure'" -o COVID-19 --expand-synonyms --lemmatise > labels.txtThis gives us some additional labels, including kidney replacement therapy for the renal replacement therapy (NCIT:C126400) class. These additional labels will allow us to recognise even more of these items in text!
Now that we have some labels in our labels.txt that describe therapeutic procedures, we can start to query EBI's PMC dataset using Komenti. For ease, we will work with this small set of labels, and only on abstracts.
$ komenti get_abstracts -l labels.txt --out abstracts/This will download the abstracts of matching articles in PubMed into the abstracts/ folder. This would give us an initial start to any manual literature review.
we can further use Komenti to associate the abstracts with the concepts we're interested in, by annotating them:
$ komenti annotate -l labels.txt -t abstracts/ --out ann.txtThis produces an ann.txt file, which is a list of every mention of each of our concepts of interest, with its surrounding sentence as context. For example:
203 PMC7291880.txt http://purl.obolibrary.org/obo/NCIT_C94624 oxygen therapy supplemental oxygen therapy isClinicalAspectFor some COVID-19 and 'therapeutic procedure' 8 most patients (23/29; 79.8 %) received supplemental oxygen therapy and antibiotics (23/29; 79.8%) in addition to traditional chinese medicines (26/29; 89.7%).
In this way, we are building assocaitions between our original complex class descriptions, and actual text.
With this command, you can produce a report, grouping articles by topics in the vocabulary:
$ komenti summarise -a ann.txtWe can further extend the functionality we have thus investigated to look at co-ocurrence between groups of concepts or complex class descriptions:
komenti summarise_entity_pairs -l labels.txt -a annotation.txt -c asbestos,toxicity