HCLSIG/PharmaOntology/Publications/JBMS-reviews

review 1: +1

This paper describes the translational medicine ontology work being carried out within the HCLS IG. It develops a framework for constructing a translational medicine knowledge base based on use cases, ontology languages (e.g., OWL and RDFS) and a set of standard semantic web technologies including triplestore and sparql. The set of questions and scenarios are well thought out. While the work is significant and represents a good start, there are still many issues that need to be addressed. It's hard to tell how well the ficticious patient records reflect the reality of those entered in the EHR system. In addition, the following issues might be relevant.

1. Security -- patient confidentiality and privacy is a main concern in patient health records.

2. Longitudinal data -- while it was mentioned the paper, it doesn't seem to be addressed in TMO. For example, how can one use the ontology to monitor patient's progression of disease?

Finally, in the first sparql example that identifies AD patients "without the apoe4 allele". The semantics in the sparql query is not clear. All it does is to perform a string match against the value of the apoe4 property.

review 2: +1

Summary The article describes the process and preliminary results of creating an ontology and knowledge base for representing clinical and biological information. The project uses standard semantic web technology; the ontology is represented in OWL, the associated knowledge base implemented in RDF and queries are conducted using SPARQL. The ontology emphasizes the description of clinical events such as would generally be recorded in an electronic medical record and puts a significant amount of effort into maintaining mappings to extant ontologies. As of today, the knowledge base is operational and the queries described in the text are working.

General comments The article describes what is clearly a work in progress. With only 7 made up patient records it is difficult to judge the value of the clinical aspect of the ontology and overall strategy proposed in the article. As discussed in more detail below, the clarity of the article could also be improved. Despite these weaknesses, the article presents strong early steps down the difficult, important and likely very long road to effective data management in the domain of translational medicine.

Specific comments

1) The text in the “Use Case” section does not seem to describe a use case. It introduces Alzheimer’s Disease and suggests that data integration is a good idea – citing the New York Times for support – but does not present the very specific actors and actions one would expect from a typical use case. It would benefit the clarity of the paper substantially if the specific use case employed here (which is referred to in a footnote in the Methods section which follows the Use Case section) was presented in the “Use Case” section of the manuscript.

2) The third sentence in the Ontology Design section points to “research questions” in Table 1, but Table 1 does not list any questions.

3) The fourth sentence in the Ontology Design section “Terms that refer to real world entities are were then represented…” has a typo (are/were). In addition, if the authors are going to raise the specter of the “real world” here in this section, then they need to specify what their criteria for reality are to make it clear to the reader what they mean by ‘real’. What are examples of terms from the texts that they used to start the ontology engineering process that do not refer to entities in the real world?

4) “Relations are specified using the Relation Ontology” should cite a reference for the R.O.

5) Figure 3 is referred to in the sentence “Given the prevalence…” as having something to do with equivalence mapping but the Figure in the current version of the manuscript seems to have nothing to do with mappings. This looks like the wrong figure was uploaded ??

6) The article would be clearer if the questions referred to in both the Unit Testing and Data Mapping sections were spelled out earlier in the text (see comment 1 again here)

7) The “SPARQL queries” section is very dependent on the existence of this footnoted link (http://esw.w3.org/HCLSIG/PharmaOntology/Queries). It might be better for the journal if these queries were included as supplementary data. (Note that all the queries are operational at the time of this review.)

8) The first sentence of the SPARQL queries section discusses 12 questions while 14 are listed.