=Paper=
{{Paper
|id=Vol-2788/om2020_poster1
|storemode=property
|title=Ontology alignment in ecotoxicological effect prediction
|pdfUrl=https://ceur-ws.org/Vol-2788/om2020_poster1.pdf
|volume=Vol-2788
|authors=Erik B. Myklebust,Ernesto Jiménez-Ruiz,Jiaoyan Chen,Raoul Wolf,Knut Erik Tollefsen
|dblpUrl=https://dblp.org/rec/conf/semweb/MyklebustJCWT20
}}
==Ontology alignment in ecotoxicological effect prediction==
https://ceur-ws.org/Vol-2788/om2020_poster1.pdf
Ontology Alignment in
Ecotoxicological Effect Prediction?
Erik B. Myklebust1,2 , Ernesto Jiménez-Ruiz2,3 , Jiaoyan Chen4 ,
Raoul Wolf1 , and Knut Erik Tollefsen1,5
1
Norwegian Institute for Water Research, Oslo, Norway
2
SIRIUS, University of Oslo, Oslo, Norway
3
City, University of London, London, United Kingdom
4
University of Oxford, Oxford, United Kingdom
5
Norwegian University of Life Sciences, Ås, Norway
1 Introduction
The Toxicological and Risk Assessment Knowledge Graph (TERA) [1] integrates sev-
eral disparate datasets relevant to ecological risk assessment and effect prediction. TERA
is being used in conjunction with knowledge graph embedding models to improve the
extrapolation of chemical effect data in the Norwegian Institute for Water Research
(Norsk institutt for vannforskning, NIVA) [1].1
The largest publicly available repository of effect data is the ECOTOXicology knowl-
edge base (ECOTOX) developed by the US Environmental Protection Agency [2]. The
dataset consists of 940k experiments using 12k compounds and 13k species. ECOTOX
contains a taxonomy (of species), however, this only considers the species represented
in the ECOTOX effect data. Hence, to enable extrapolation of effects across a larger tax-
onomic domain, an alignment to the NCBI taxonomy have to be established. However,
there does not exist a complete and public mapping set between the 47,785 ECOTOX
taxa and the 2,140,344 NCBI taxa. In this paper we present the ECOTOX-NCBI align-
ment results of three ontology matching algorithms.
2 Methods and Evaluation
Although there does not exist a complete and public alignment between the ECOTOX
and NCBI, a partial mapping curated by experts can be obtained through the ECOTOX
Web.2 We have gathered a total of 2,321 mappings for validation purposes. We have
used three methods to align the two vocabularies: (i) LogMap system [3]. (ii) Agree-
mentMakerLight (AML) , and (iii) a baseline string matching algorithm based on Lev-
enshtein distance [4].
Table 1 shows the alignment results over the ground truth samples. Note that the
results represent 1-to-1 alignments as, in our setting, it is expected an entity from
?
Copyright c 2020 for this paper by its authors. Use permitted under Creative Commons Li-
cense Attribution 4.0 International (CC BY 4.0).
1
Knowledge Graphs at NIVA: https://github.com/NIVA-Knowledge-Graph/
2
ECOTOX search interface: https://cfpub.epa.gov/ecotox/search.cfm
Algorithm # mappings Recall Precision (*)
LogMap 32, 726 0.81 0.88
AML 31, 659 0.80 0.87
String distance (> 0.8) 33, 554 0.38 0.70
Union all 57, 511 0.72 0.73
Consensus (LogMap ∩ AML) 20, 217 0.78 0.95
LogMap ∪ AML 39, 985 0.83 0.85
Table 1. Alignment results for ECOTOX-NCBI. (*) Estimated precision with respect to the
known entities in the incomplete reference alignment, assuming only 1-1 mappings are valid.
ECOTOX to match to a single entity in NCBI, and vice-versa. Hence, 1-to-N (respec-
tively N-to-1) alignments were filtered according to the system computed confidence.
LogMap and AML produce mapping sets with similar recall and (estimated) precision,
with LogMap producing a larger number of mappings. The baseline matcher, as ex-
pected, achieves both a lower recall and (estimated) precision. This shows that a simple
string matching solution may not be enough in this setting. Table 1 also shows the re-
sults of the consensus alignment between AML and LogMap and the union of different
mapping sets. Note that the lower recall of the union is down to overconfidence in the
string distance method when 1-to-1 filtering.
3 Conclusions
The used alignment techniques achieve relatively good scores for recall over the avail-
able (incomplete) reference mappings. However, aligning such large and challenging
datasets required some preprocessing before ontology alignment systems could cope
with them. The preprocessing involved to split NCBI into manageable fragments, lead-
ing to a set of matching subtasks instead of a single task. Thus, the alignment of ECO-
TOX and NCBI has the potential of becoming a new track of the Ontology Alignment
Evaluation Initiative (OAEI)3 [5] to push the limits of state-of-the-art systems. The
output of the different OAEI participants could be merged into a rich consensus align-
ment that could become the reference to integrate ECOTOX and NCBI. At the same
time, as the alignment between ECOTOX and NCBI is not public nor complete, the
consensus mappings could also be seen as a very relevant resource to the ecotoxicology
community.
References
1. Myklebust, E.B., Jiménez-Ruiz, E., Chen, J., Wolf, R., Tollefsen, K.E.: Knowledge Graph
Embedding for Ecotoxicological Effect Prediction. In: ISWC. (2019)
2. U.S. EPA: ECOTOXicology knowledgebase (ECOTOX) (2019)
3. Jiménez-Ruiz, E., Cuenca Grau, B., Zhou, Y., Horrocks, I.: Large-scale interactive ontology
matching: Algorithms and implementation. In: ECAI. (2012)
4. Levenshtein, V.I.: Binary Codes Capable of Correcting Deletions, Insertions and Reversals.
Soviet Physics Doklady 10 (1966)
5. Algergawy, A., et al.: Results of the Ontology Alignment Evaluation Initiative 2019. In: 14th
International Workshop on Ontology Matching. (2019) 46–85
3
OAEI: http://oaei.ontologymatching.org/