pith. machine review for the scientific record. sign in

arxiv: 2605.03690 · v1 · submitted 2026-05-05 · 💻 cs.LG · cs.AI· q-bio.QM

Recognition: unknown

Graph Neural Network based Hierarchy-Aware Embeddings of Knowledge Graphs: Applications to Yeast Phenotype Prediction

Filip Kronstr\"om , Alexander H. Gower , Daniel Brunns{\aa}ker , Ievgeniia A. Tiukova , Ross D. King
Authors on Pith no claims yet

Pith reviewed 2026-05-07 16:58 UTC · model grok-4.3

classification 💻 cs.LG cs.AIq-bio.QM
keywords graph neural networksknowledge graphsembeddingsyeastphenotype predictionsemantic lossontologygene knockouts
0
0 comments X

The pith

Graph neural networks enriched with semantic loss produce hierarchy-aware embeddings of yeast knowledge graphs that predict cell growth phenotypes from gene knockouts.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces a method to create embeddings of knowledge graphs that respect ontological hierarchies by training graph neural networks with an additional semantic loss term. This is demonstrated on a knowledge graph of yeast biology to predict the outcomes of gene deletions on cell growth. The predictions achieve a mean R-squared score of 0.360 across cross-validation, which improves to 0.377 when the semantic loss is used, outperforming baselines. This shows that high-level qualitative knowledge from ontologies can inform quantitative experimental predictions, and the approach also generates hypotheses about gene interactions that were validated in a lab experiment.

Core claim

Low-dimensional box embeddings of the yeast knowledge graph combined with graph neural networks can predict cell growth for double gene knockouts with a mean R² score of 0.360 over 10-fold cross validation, significantly higher than baselines. Incorporating semantic loss terms improves performance to R²=0.377 by aligning the embeddings with the structure of the ontology. The models generalize to triple gene knockouts, and analysis of important relations in the graph leads to hypotheses about interacting traits, one of which was confirmed experimentally as an association between inositol utilisation and osmotic stress resistance.

What carries the argument

Hierarchy-aware box embeddings learned by graph neural networks augmented with a semantic loss function derived from ontology class hierarchies.

If this is right

  • The predictions generalize from double to triple gene knockouts, indicating the model captures broader interaction patterns.
  • Analysis of co-occurring relations in the knowledge graph can generate testable hypotheses about biological traits in yeast.
  • Box embeddings can be used to evaluate the quality of revisions to the knowledge graph.
  • Semantic loss improves alignment with ontology structure, leading to better predictive performance.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The method could extend to predicting phenotypes in other organisms if similar knowledge graphs and ontologies are available.
  • Incorporating ontology hierarchies this way might improve embedding-based models in other scientific domains with structured knowledge.
  • Validated hypotheses from such models could accelerate biological discovery by prioritizing experiments.
  • Box embeddings might help detect inconsistencies or gaps in existing knowledge graphs beyond the yeast example.

Load-bearing premise

The yeast knowledge graph built from community databases accurately captures the biologically relevant relationships that determine cell-growth phenotypes under gene deletions.

What would settle it

If the mean R² score on a new set of triple gene knockout experiments or other held-out data falls to the level of baseline models without semantic loss, the claim that the embeddings improve predictions would be falsified.

Figures

Figures reproduced from arXiv: 2605.03690 by Alexander H. Gower, Daniel Brunns{\aa}ker, Filip Kronstr\"om, Ievgeniia A. Tiukova, Ross D. King.

Figure 1
Figure 1. Figure 1: Overview of the Hierarchy-aware GNN. a shows how the output of each message passing layer, which aggregates information between neighbours in the KG, is treated as a latent variable that is converted into boxes through the box trans￾formation in (6). The boxes are trained to fulfil specified class hierarchies using the losses in (10-13), which can also be applied to the prior node embedding. The output of … view at source ↗
Figure 2
Figure 2. Figure 2: An overview of the different types of classes and how they are connected in the knowledge graph is shown in a. The colour of the nodes specifies where the classes are defined. b shows examples from the hierarchies defining classes in the domains introduced in Section 5.3 view at source ↗
Figure 3
Figure 3. Figure 3: An overview of the system predicting the fitness when deleting pairs of genes is shown in a. A GNN using GraphSAGE message passing layers, acting on the KG from Section 5.2, generates node embeddings. The embeddings of pairs of genes are combined through element-wise multiplication and fed to a NN predicting the fitness of the gene deletion. b shows how classes in the different domains are represented by b… view at source ↗
Figure 4
Figure 4. Figure 4: Parity plots for double, (a) and (c), and triple, (b) and (d), gene deletions. (a) and (b) shows the parity plot for the model using box embeddings as prior node representations only, while (c) and (d) shows the predictions from a model also trained with the distance-based semantic losses, Ldistance. For the double dele￾tion, the predictions from all validation sets in the cross validation are shown. ure 4… view at source ↗
Figure 5
Figure 5. Figure 5: Average Ldistance and L − distance losses per class for the different domains in the KG, during training of the best performing model (f) in view at source ↗
Figure 6
Figure 6. Figure 6: An overview of the selection and results of the experiment we performed. (a) shows the highest ranked importances of edge-pairs and the pair selected for the experiment, nutrient utilisation of inositol and stress resistance to NaCl, is highlighted in red. f0 and f1, which have a higher assigned weight, are discarded due to safety and lab constraints as it involves the chemical bleomycin. (b) Box plot show… view at source ↗
Figure 7
Figure 7. Figure 7: Learned box embeddings in two dimensions for the molecular function domain. 7(a) and 7(b) box embeddings prior to input into GNN; 7(b) and 7(d) show final embeddings for distance and overlap loss respectively. The main advantage of our approach is that signals from semantic information encoded in class hierarchies, and signals from predictive tasks can jointly be used to train graph neural networks. We dem… view at source ↗
Figure 8
Figure 8. Figure 8: Distribution of distances of box embeddings learned from revised graphs G˜ to the original embeddings learned from Gtrain, shown by relation type, for a subset of the relation types in the graph. (The method for calculating these differences is described in 5.6). For most relations, the distance ranks of randomly drawn edges (constrained to the appropriate class) were significantly different to the ranks o… view at source ↗
Figure 9
Figure 9. Figure 9: Growth curves showing the mean optical densities of the 6-8 repetitions for the different experimental groups. Optical density (at 600nM) is a unitless measure￾ment typically used as an indirect measure of cell density and biomass. 48 view at source ↗
read the original abstract

We present a method for finding hierarchy-aware embeddings of knowledge graphs (KGs) using graph neural networks (GNNs) enriched with a semantic loss derived from underlying ontologies. This method yields embeddings that better reflect domain knowledge. To demonstrate their utility, we predict and interpret the effects of gene deletions in the yeast Saccharomyces cerevisiae and learn box embeddings for KGs in the absence of a prediction task. We further show how box embeddings can serve as the basis for evaluating KG revisions. Our yeast KG is constructed from community databases and ontology terms. Low-dimensional box embeddings combined with GNNs are used to predict cell growth for double gene knockouts. Over 10-fold cross validation, these predictions have a mean $R^2$~score~of~0.360, significantly higher than baseline comparisons, demonstrating that high-level qualitative knowledge is informative about experimental outcomes. Incorporating semantic loss terms in the training of the models improves their predictive performance ($R^2$=0.377) by aligning embeddings with ontology structure. This shows that class hierarchies from ontologies can be exploited for quantitative prediction. We also test the trained models on triple gene knockouts, showing they generalise to data beyond those seen in training. Additionally, by identifying co-occurring relations in the yeast KG important for the cell-growth predictions, we construct hypotheses about interacting traits in yeast. A biological experiment validates one such finding, revealing an association between inositol utilisation and osmotic stress resistance, highlighting the model's potential to guide biological discovery.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces a GNN-based approach to learning hierarchy-aware box embeddings for knowledge graphs by adding a semantic loss term derived from ontology class hierarchies. The method is applied to a yeast KG built from public databases and ontologies to predict cell-growth phenotypes under double gene knockouts, reporting a mean R² of 0.360 over 10-fold cross-validation that rises to 0.377 with the semantic loss; the models are further tested on triple knockouts, used to evaluate KG revisions, and employed to generate hypotheses that receive one biological validation.

Significance. If the reported R² gains are free of transductive leakage, the work provides concrete evidence that ontological hierarchy information can be turned into quantitative predictive gains for experimental phenotypes in a model organism. The biological validation experiment and the generalization test on triple knockouts are positive features that strengthen the utility claim. The semantic-loss construction and box-embedding evaluation of KG revisions may be reusable in other hierarchical domains.

major comments (2)
  1. [cross-validation procedure and embedding training details] The description of the 10-fold cross-validation for double-knockout phenotype prediction (abstract and experimental results section) does not state whether the GNN parameters and semantic loss are recomputed from scratch on each training fold or learned once on the full yeast KG. If the latter, every gene participates in message passing and ontology alignment even when its knockout pairs are held out, creating a transductive setting in which test-gene representations can encode information from training phenotypes via shared neighbors or ontology paths. This directly affects the central claim that the R² improvement from 0.360 to 0.377 is attributable to hierarchy alignment rather than leakage.
  2. [baseline construction in experimental evaluation] The baseline comparisons that yield the reported R² of 0.360 are not described in sufficient detail (results section) to allow assessment of whether they use the same KG structure, the same box-embedding parameterization, or the same feature construction; without this, the magnitude of the improvement cannot be interpreted as evidence that the semantic loss adds value beyond standard GNN message passing.
minor comments (2)
  1. [methods] Notation for the semantic loss term and the box-embedding distance function should be introduced with explicit equations rather than prose descriptions only.
  2. [data description] The manuscript would benefit from a table listing the exact data sources, number of nodes/edges/relations in the yeast KG, and the phenotype dataset size and split statistics.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed review. We address each major comment below and will revise the manuscript to improve clarity on the experimental procedures.

read point-by-point responses

  1. Referee: The description of the 10-fold cross-validation for double-knockout phenotype prediction (abstract and experimental results section) does not state whether the GNN parameters and semantic loss are recomputed from scratch on each training fold or learned once on the full yeast KG. If the latter, every gene participates in message passing and ontology alignment even when its knockout pairs are held out, creating a transductive setting in which test-gene representations can encode information from training phenotypes via shared neighbors or ontology paths. This directly affects the central claim that the R² improvement from 0.360 to 0.377 is attributable to hierarchy alignment rather than leakage.

    Authors: We acknowledge that the manuscript did not explicitly describe the cross-validation procedure in sufficient detail. The GNN parameters and semantic loss are computed once on the full yeast KG (including all genes and ontology paths), after which the 10-fold CV is performed solely on the downstream double-knockout phenotype prediction task. This follows the standard transductive protocol for KG embedding methods, where the complete graph structure is used to learn representations that capture relational and hierarchical information. Both the baseline model (mean R² = 0.360) and the semantic-loss model (mean R² = 0.377) employ identical embedding training on the full KG, so the reported improvement isolates the contribution of the ontology-derived semantic loss. We will revise the experimental results section to state this procedure explicitly, include a discussion of the transductive setting, and clarify that the KG structure itself is label-independent. revision: yes

  2. Referee: The baseline comparisons that yield the reported R² of 0.360 are not described in sufficient detail (results section) to allow assessment of whether they use the same KG structure, the same box-embedding parameterization, or the same feature construction; without this, the magnitude of the improvement cannot be interpreted as evidence that the semantic loss adds value beyond standard GNN message passing.

    Authors: We agree that the baseline methods require additional implementation details. The baselines are standard GNN models without the semantic loss term; they operate on the identical yeast KG, employ the same box-embedding parameterization, and use the same node feature construction and training protocol as the proposed method. This design isolates the effect of the semantic loss. We will expand the results section in the revised manuscript with a dedicated subsection describing the baseline architectures, hyperparameters, and training procedures to ensure full reproducibility and allow readers to verify the fairness of the comparison. revision: yes

Circularity Check

0 steps flagged

No circularity: embeddings and semantic loss derive from external KG and ontologies; predictions evaluated on held-out experimental phenotypes

full rationale

The paper builds the yeast KG from community databases and ontology terms separate from the phenotype labels. It trains GNN box embeddings with a semantic loss that aligns to external ontology hierarchies, then measures R² on 10-fold CV held-out double-knockout growth data (and generalizes to triples). No derivation step equates the reported performance gain or hierarchy alignment to a quantity defined solely by the fitted parameters, self-citations, or input data by construction. The improvement is an empirical outcome on independent experimental targets, not a tautology.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that the constructed yeast KG faithfully encodes relevant biology and that semantic loss from ontology hierarchies can be defined without additional ad-hoc parameters; no free parameters or invented entities are explicitly introduced in the abstract.

axioms (1)
  • domain assumption The yeast KG built from community databases and ontologies accurately captures relationships relevant to cell growth phenotypes.
    Invoked when claiming that high-level qualitative knowledge is informative about experimental outcomes.

pith-pipeline@v0.9.0 · 5606 in / 1282 out tokens · 60699 ms · 2026-05-07T16:58:24.121907+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

95 extracted references · 57 canonical work pages · 1 internal anchor

  1. [2]

    An Introduction to Description Logic

    Franz Baader, Ian Horrocks, Carsten Lutz, and Uli Sattler. An Introduction to Description Logic. Cambridge University Press. ISBN 978-0-521-87361-1. doi:10.1017/9781139025355

    work page doi:10.1017/9781139025355

  2. [3]

    Translating embeddings for modeling multi-relational data

    Antoine Bordes, Nicolas Usunier, Alberto Garcia-Durán, Jason Weston, and Oksana Yakhnenko. Translating embeddings for modeling multi-relational data. In Proceedings of the 27th International Conference on Neural Information Processing Systems - Volume 2, volume 2 of NIPS '13 , pages 2787--2795. Curran Associates Inc., 2013

    2013

  3. [5]

    Gower, Prajakta Naval, Erik Y

    Daniel Brunnsåker, Alexander H. Gower, Prajakta Naval, Erik Y. Bjurström, Filip Kronström, Ievgeniia A. Tiukova, and Ross D. King. Agentic AI integrated with scientific knowledge: Laboratory validation in systems biology, 2025. URL https://www.biorxiv.org/content/10.1101/2025.06.24.661378v3. ISSN : 2692-8205 Pages: 2025.06.24.661378 Section: New Results

    work page doi:10.1101/2025.06.24.661378v3 2025

  4. [6]

    Box embeddings: An open-source library for representation learning using geometric structures, 2021

    Tejas Chheda, Purujit Goyal, Trang Tran, Dhruvesh Patel, Michael Boratko, Shib Sankar Dasgupta, and Andrew McCallum . Box embeddings: An open-source library for representation learning using geometric structures, 2021. URL http://arxiv.org/abs/2109.04997

    work page arXiv 2021

  5. [7]

    Le, and Christopher D

    Kevin Clark, Minh-Thang Luong, Quoc V. Le, and Christopher D. Manning. ELECTRA : Pre-training text encoders as discriminators rather than generators. 2019. URL https://openreview.net/forum?id=r1xMH1BtvB

    2019

  6. [12]

    Improving local identifiability in probabilistic box embeddings

    Shib Sankar Dasgupta, Michael Boratko, Dongxu Zhang, Luke Vilnis, Xiang Lorraine Li, and Andrew McCallum . Improving local identifiability in probabilistic box embeddings. In Proceedings of the 34th International Conference on Neural Information Processing Systems, NIPS '20, pages 182--192. Curran Associates Inc., 2020. ISBN 978-1-7138-2954-6

    2020

  7. [15]

    Hyperbolic entailment cones for learning hierarchical embeddings

    Octavian Ganea, Gary Becigneul, and Thomas Hofmann. Hyperbolic entailment cones for learning hierarchical embeddings. In Proceedings of the 35th International Conference on Machine Learning, pages 1646--1655. PMLR , 2018. ISSN : 2640-3498

    2018

  8. [17]

    Schockaert

    Víctor Gutiérrez-Basulto and S. Schockaert. From knowledge graph embedding to ontology embedding? A n analysis of the compatibility between vector space representations and rules. In International Conference on Principles of Knowledge Representation and Reasoning, 2018

    2018

  9. [18]

    Hamilton, Rex Ying, and Jure Leskovec

    William L. Hamilton, Rex Ying, and Jure Leskovec. Inductive representation learning on large graphs. In Proceedings of the 31st International Conference on Neural Information Processing Systems, NIPS '17, pages 1025--1035. Curran Associates Inc., 2017. ISBN 978-1-5108-6096-4

    2017

  10. [21]

    Open graph benchmark: datasets for machine learning on graphs

    Weihua Hu, Matthias Fey, Marinka Zitnik, Yuxiao Dong, Hongyu Ren, Bowen Liu, Michele Catasta, and Jure Leskovec. Open graph benchmark: datasets for machine learning on graphs. In Proceedings of the 34th International Conference on Neural Information Processing Systems, NIPS '20, pages 22118--22133. Curran Associates Inc., 2020. ISBN 978-1-7138-2954-6

    2020

  11. [25]

    LightGBM : A highly efficient gradient boosting decision tree

    Guolin Ke, Qi Meng, Thomas Finley, Taifeng Wang, Wei Chen, Weidong Ma, Qiwei Ye, and Tie-Yan Liu. LightGBM : A highly efficient gradient boosting decision tree. In Advances in Neural Information Processing Systems, volume 30. Curran Associates, Inc., 2017. URL https://papers.nips.cc/paper_files/paper/2017/hash/6449f44a102fde848669bdd9eb6b76fa-Abstract.html

    2017

  12. [27]

    Kipf and Max Welling

    Thomas N. Kipf and Max Welling. Semi-supervised classification with graph convolutional networks, 2017

    2017

  13. [28]

    Captum: A unified and generic model interpretability library for pytorch, 2020

    Narine Kokhlikyan, Vivek Miglani, Miguel Martin, Edward Wang, Bilal Alsallakh, Jonathan Reynolds, Alexander Melnikov, Natalia Kliushkina, Carlos Araya, Siqi Yan, and Orion Reblitz-Richardson. Captum: A unified and generic model interpretability library for pytorch, 2020

    2020

  14. [29]

    Tiukova, and Ross D

    Filip Kronstr \"o m, Daniel Brunns a ker, Ievgeniia A. Tiukova, and Ross D. King. Ontology-based box embeddings and knowledge graphs for predicting phenotypic traits in saccharomyces cerevisiae. In 19th International Conference on Neurosymbolic Learning and Reasoning, 2025

    2025

  15. [38]

    Chris Mungall, David Osumi-Sutherland, James A. Overton, Jim Balhoff, Clare72, pgaudet, Matthew Brush, Nico Matentzoglu, Vasundra Touré, Damion Dooley, Michael Sinclair, Anthony Bretaudeau, Scott Cain, Melissa Haendel, diatomsRcool , Jen Hammock, Marie-Angélique Laporte, Mark Jensen, and Martin Larralde. oborel/obo-relations: 2020-07-21, 2020. URL https:/...

    work page arXiv 2020

  16. [39]

    Poincaré embeddings for learning hierarchical representations

    Maximilian Nickel and Douwe Kiela. Poincaré embeddings for learning hierarchical representations. In Proceedings of the 31st International Conference on Neural Information Processing Systems, NIPS '17, pages 6341--6350. Curran Associates Inc., 2017. ISBN 978-1-5108-6096-4

    2017

  17. [40]

    A three-way model for collective learning on multi-relational data

    Maximilian Nickel, Volker Tresp, and Hans-Peter Kriegel. A three-way model for collective learning on multi-relational data. In Proceedings of the 28th International Conference on International Conference on Machine Learning, ICML '11, pages 809--816. Omnipress, 2011. ISBN 978-1-4503-0619-5

    2011

  18. [42]

    Representing joint hierarchies with box embeddings

    Dhruvesh Patel, Shib Sankar Dasgupta, Michael Boratko, Xiang Li, Luke Vilnis, and Andrew McCallum. Representing joint hierarchies with box embeddings. In Automated Knowledge Base Construction ( AKBC ) , 2020. doi:10.24432/C5KS37

    work page doi:10.24432/c5ks37 2020

  19. [43]

    Description logic EL ++ embeddings with intersectional closure, 2022

    Xi Peng, Zhenwei Tang, Maxat Kulmanov, Kexin Niu, and Robert Hoehndorf. Description logic EL ++ embeddings with intersectional closure, 2022. URL http://arxiv.org/abs/2202.14018

    work page arXiv 2022

  20. [46]

    Not just a black box: Learning important features through propagating activation differences, 2017

    Avanti Shrikumar, Peyton Greenside, Anna Shcherbina, and Anshul Kundaje. Not just a black box: Learning important features through propagating activation differences, 2017. URL http://arxiv.org/abs/1605.01713

    work page arXiv 2017

  21. [47]

    RotatE : Knowledge graph embedding by relational rotation in complex space

    Zhiqing Sun, Zhi-Hong Deng, Jian-Yun Nie, and Jian Tang. RotatE : Knowledge graph embedding by relational rotation in complex space. In 7th International Conference on Learning Representations, ICLR 2019, New Orleans, LA , USA , May 6-9, 2019 . OpenReview .net, 2019

    2019

  22. [48]

    Complex embeddings for simple link prediction

    Théo Trouillon, Johannes Welbl, Sebastian Riedel, Eric Gaussier, and Guillaume Bouchard. Complex embeddings for simple link prediction. In Proceedings of The 33rd International Conference on Machine Learning, pages 2071--2080. PMLR , 2016. ISSN : 1938-7228

    2071

  23. [50]

    Probabilistic embedding of knowledge graphs with box lattice measures

    Luke Vilnis, Xiang Li, Shikhar Murty, and Andrew McCallum . Probabilistic embedding of knowledge graphs with box lattice measures. In Iryna Gurevych and Yusuke Miyao, editors, Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 263--272. Association for Computational Linguistics, 2018. doi...

    work page doi:10.18653/v1/p18-1025 2018

  24. [55]

    A semantic loss function for deep learning with symbolic knowledge

    Jingyi Xu, Zilu Zhang, Tal Friedman, Yitao Liang, and Guy Van den Broeck. A semantic loss function for deep learning with symbolic knowledge. In Jennifer Dy and Andreas Krause, editors, Proceedings of the 35th International Conference on Machine Learning, volume 80 of Proceedings of Machine Learning Research, pages 5502--5511. PMLR, 10--15 Jul 2018

    2018

  25. [56]

    Embedding entities and relations for learning and inference in knowledge bases

    Bishan Yang, Wen-tau Yih, Xiaodong He, Jianfeng Gao, and Li Deng. Embedding entities and relations for learning and inference in knowledge bases. In Yoshua Bengio and Yann LeCun , editors, 3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA , USA , May 7-9, 2015, Conference Track Proceedings , 2015

    2015

  26. [60]

    and Aleksander, Suzi and Nash, Robert S

    Engel, Stacia R. and Aleksander, Suzi and Nash, Robert S. and Wong, Edith D. and Weng, Shuai and Miyasato, Stuart R. and Sherlock, Gavin and Cherry, J. Michael , date =. Saccharomyces Genome Database: Advances in Genome Annotation, Expanded Biochemical Pathways, and Other Key Enhancements , issn =. Genetics , year =. doi:10.1093/genetics/iyae185 , shorttitle =

    work page doi:10.1093/genetics/iyae185

  27. [61]

    Ball, Judith A

    Ashburner, Michael and Ball, Catherine A. and Blake, Judith A. and Botstein, David and Butler, Heather and Cherry, J. Michael and Davis, Allan P. and Dolinski, Kara and Dwight, Selina S. and Eppig, Janan T. and Harris, Midori A. and Hill, David P. and Issel-Tarver, Laurie and Kasarskis, Andrew and Lewis, Suzanna and Matese, John C. and Richardson, Joel E....

    work page doi:10.1038/75556 2000

  28. [62]

    Mungall, Chris and Osumi-Sutherland, David and Overton, James A. and Balhoff, Jim and Clare72 and pgaudet and Brush, Matthew and Matentzoglu, Nico and Touré, Vasundra and Dooley, Damion and Sinclair, Michael and Bretaudeau, Anthony and Cain, Scott and Haendel, Melissa and. oborel/obo-relations: 2020-07-21 , url =. 2020 , doi =

    2020

  29. [63]

    and Skrzypek, Marek S

    Costanzo, Maria C. and Skrzypek, Marek S. and Nash, Robert and Wong, Edith and Binkley, Gail and Engel, Stacia R. and Hitz, Benjamin and Hong, Eurie L. and Cherry, J. Michael and. New mutant phenotype data curation system in the Saccharomyces Genome Database , volume =. 2009 , journal =. doi:10.1093/database/bap001 , abstract =

    work page doi:10.1093/database/bap001 2009

  30. [64]

    Nucleic acids research , shortjournal =

    Hastings, Janna and Owen, Gareth and Dekker, Adriano and Ennis, Marcus and Kale, Namrata and Muthukrishnan, Venkatesh and Turner, Steve and Swainston, Neil and Mendes, Pedro and Steinbeck, Christoph , urldate =. Nucleic acids research , shortjournal =. 2016 , pmid =. doi:10.1093/nar/gkv1031 , shorttitle =

    work page doi:10.1093/nar/gkv1031 2016

  31. [65]

    Development and application of an interaction network ontology for literature mining of vaccine-associated gene-gene interactions , volume =

    Hur, Junguk and Özgür, Arzucan and Xiang, Zuoshuang and He, Yongqun , urldate =. Development and application of an interaction network ontology for literature mining of vaccine-associated gene-gene interactions , volume =. Journal of Biomedical Semantics , shortjournal =. 2015 , keywords =. doi:10.1186/2041-1480-6-2 , abstract =

    work page doi:10.1186/2041-1480-6-2 2015

  32. [66]

    and Billington, Richard and Caspi, Ron and Fulcher, Carol A

    Karp, Peter D. and Billington, Richard and Caspi, Ron and Fulcher, Carol A. and Latendresse, Mario and Kothari, Anamika and Keseler, Ingrid M. and Krummenacker, Markus and Midford, Peter E. and Ong, Quang and Ong, Wai Kit and Paley, Suzanne M. and Subhraveti, Pallavi , date =. The. Briefings in Bioinformatics , shortjournal =. 2019 , pmid =. doi:10.1093/b...

    work page doi:10.1093/bib/bbx085 2019

  33. [67]

    and Rutherford, Kim and Bähler, Jürg and Oliver, Stephen G

    Wood, Valerie and Lock, Antonia and Harris, Midori A. and Rutherford, Kim and Bähler, Jürg and Oliver, Stephen G. , urldate =. Hidden in plain sight: what remains to be discovered in the eukaryotic proteome? , volume =. Open Biology , shortjournal =. 2019 , date =. doi:10.1098/rsob.180241 , shorttitle =

    work page doi:10.1098/rsob.180241 2019

  34. [68]

    Global Genetic Networks and the Genotype-to-Phenotype Relationship , volume =

    Costanzo, Michael and Kuzmin, Elena and van Leeuwen, Jolanda and Mair, Barbara and Moffat, Jason and Boone, Charles and Andrews, Brenda , urldate =. Global Genetic Networks and the Genotype-to-Phenotype Relationship , volume =. Cell , year =. doi:10.1016/j.cell.2019.01.033 , abstract =

    work page doi:10.1016/j.cell.2019.01.033 2019

  35. [69]

    Saccharomyces cerevisiae and its industrial applications , volume =

    Parapouli, Maria and Vasileiadis, Anastasios and Afendra, Amalia-Sofia and Hatziloukas, Efstathios , urldate =. Saccharomyces cerevisiae and its industrial applications , volume =. 2020 , shortjournal =. doi:10.3934/microbiol.2020001 , abstract =

    work page doi:10.3934/microbiol.2020001 2020

  36. [70]

    Nucleic Acids Research , shortjournal =

    Dubreuil, Benjamin and Sass, Ehud and Nadav, Yotam and Heidenreich, Meta and Georgeson, Joseph M and Weill, Uri and Duan, Yuanqiang and Meurer, Matthias and Schuldiner, Maya and Knop, Michael and Levy, Emmanuel D , urldate =. Nucleic Acids Research , shortjournal =. 2019 , pmid =. doi:10.1093/nar/gky941 , shorttitle =

    work page doi:10.1093/nar/gky941 2019

  37. [71]

    and Karu, Naama and Djoumbou Feunang, Yannick and Arndt, David and Wishart, David S

    Ramirez-Gaona, Miguel and Marcu, Ana and Pon, Allison and Guo, An Chi and Sajed, Tanvir and Wishart, Noah A. and Karu, Naama and Djoumbou Feunang, Yannick and Arndt, David and Wishart, David S. , date =. Nucleic Acids Research , shortjournal =. 2017 , pmid =. doi:10.1093/nar/gkw1058 , shorttitle =

    work page doi:10.1093/nar/gkw1058 2017

  38. [72]

    Buzzao, Davide and Persson, Emma and Guala, Dimitri and Sonnhammer, Erik L. L. , date =. Nucleic Acids Research , shortjournal =. 2025 , pmid =. doi:10.1093/nar/gkae1021 , shorttitle =

    work page doi:10.1093/nar/gkae1021 2025

  39. [73]

    and Grant, Seth G

    Hermjakob, Henning and Montecchi-Palazzi, Luisa and Bader, Gary and Wojcik, Jérôme and Salwinski, Lukasz and Ceol, Arnaud and Moore, Susan and Orchard, Sandra and Sarkans, Ugis and von Mering, Christian and Roechert, Bernd and Poux, Sylvain and Jung, Eva and Mersch, Henning and Kersey, Paul and Lappe, Michael and Li, Yixue and Zeng, Rong and Rana, Debashi...

    work page doi:10.1038/nbt926 2004

  40. [74]

    , date =

    Arp, Robert and Smith, Barry and Spear, Andrew D. , date =. Building Ontologies with Basic Formal Ontology , isbn =

  41. [75]

    and Nováček, Vít , urldate =

    Walsh, Brian and Mohamed, Sameh K. and Nováček, Vít , urldate =. Proceedings of the 29th. 2020 , file =. doi:10.1145/3340531.3412776 , series =

    work page doi:10.1145/3340531.3412776 2020

  42. [76]

    The scalable precision medicine open knowledge engine (

    Morris, John H and Soman, Karthik and Akbas, Rabia E and Zhou, Xiaoyuan and Smith, Brett and Meng, Elaine C and Huang, Conrad C and Cerono, Gabriel and Schenk, Gundolf and Rizk-Jackson, Angela and Harroud, Adil and Sanders, Lauren and Costes, Sylvain V and Bharat, Krish and Chakraborty, Arjun and Pico, Alexander R and Mardirossian, Taline and Keiser, Mich...

    work page doi:10.1093/bioinformatics/btad080

  43. [77]

    Open graph benchmark: datasets for machine learning on graphs , isbn =

    Hu, Weihua and Fey, Matthias and Zitnik, Marinka and Dong, Yuxiao and Ren, Hongyu and Liu, Bowen and Catasta, Michele and Leskovec, Jure , urldate =. Open graph benchmark: datasets for machine learning on graphs , isbn =. Proceedings of the 34th International Conference on Neural Information Processing Systems , publisher =. 2020 , file =

    2020

  44. [78]

    Translating embeddings for modeling multi-relational data , volume =

    Bordes, Antoine and Usunier, Nicolas and Garcia-Durán, Alberto and Weston, Jason and Yakhnenko, Oksana , urldate =. Translating embeddings for modeling multi-relational data , volume =. Proceedings of the 27th International Conference on Neural Information Processing Systems - Volume 2 , publisher =

  45. [79]

    Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence , publisher =

    Kulmanov, Maxat and Liu-Wei, Wang and Yan, Yuan and Hoehndorf, Robert , urldate =. Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence , publisher =. 2019 , langid =. doi:10.24963/ijcai.2019/845 , shorttitle =

    work page doi:10.24963/ijcai.2019/845 2019

  46. [80]

    Description logic EL ++ embeddings with intersectional closure, 2022

    Peng, Xi and Tang, Zhenwei and Kulmanov, Maxat and Niu, Kexin and Hoehndorf, Robert , urldate =. Description Logic. 2022 , eprinttype =. doi:10.48550/arXiv.2202.14018 , abstract =. 2202.14018 [cs] , keywords =

    work page doi:10.48550/arxiv.2202.14018 2022

  47. [81]

    Improving local identifiability in probabilistic box embeddings , isbn =

    Dasgupta, Shib Sankar and Boratko, Michael and Zhang, Dongxu and Vilnis, Luke and Li, Xiang Lorraine and. Improving local identifiability in probabilistic box embeddings , isbn =. Proceedings of the 34th International Conference on Neural Information Processing Systems , publisher =. 2020 , file =

    2020

  48. [82]

    and Ying, Rex and Leskovec, Jure , urldate =

    Hamilton, William L. and Ying, Rex and Leskovec, Jure , urldate =. Inductive representation learning on large graphs , isbn =. Proceedings of the 31st International Conference on Neural Information Processing Systems , publisher =. 2017 , file =

    2017

  49. [83]

    Using deep learning to model the hierarchical structure and function of a cell , volume =

    Ma, Jianzhu and Yu, Michael Ku and Fong, Samson and Ono, Keiichiro and Sage, Eric and Demchak, Barry and Sharan, Roded and Ideker, Trey , urldate =. Using deep learning to model the hierarchical structure and function of a cell , volume =. Nature Methods , shortjournal =. 2018 , langid =. doi:10.1038/nmeth.4627 , abstract =

    work page doi:10.1038/nmeth.4627 2018

  50. [84]

    Interpreting protein abundance in Saccharomyces cerevisiae through relational learning , volume =

    Brunnsåker, Daniel and Kronström, Filip and Tiukova, Ievgeniia A and King, Ross D , urldate =. Interpreting protein abundance in Saccharomyces cerevisiae through relational learning , volume =. Bioinformatics , shortjournal =. 2024 , file =. doi:10.1093/bioinformatics/btae050 , abstract =

    work page doi:10.1093/bioinformatics/btae050 2024

  51. [85]

    Predicting gene disease associations with knowledge graph embeddings for diseases with curtailed information , volume =

    Gualdi, Francesco and Oliva, Baldomero and Piñero, Janet , urldate =. Predicting gene disease associations with knowledge graph embeddings for diseases with curtailed information , volume =. 2024 , file =. doi:10.1093/nargab/lqae049 , abstract =

    work page doi:10.1093/nargab/lqae049 2024

  52. [86]

    Box embeddings: An open-source library for representation learning using geometric structures, 2021

    Chheda, Tejas and Goyal, Purujit and Tran, Trang and Patel, Dhruvesh and Boratko, Michael and Dasgupta, Shib Sankar and. Box Embeddings: An open-source library for representation learning using geometric structures , url =. 2021 , langid =. doi:10.48550/arXiv.2109.04997 , shorttitle =. 2109.04997 [cs] , keywords =

    work page doi:10.48550/arxiv.2109.04997 2021

  53. [87]

    Science , author =

    A global genetic interaction network maps a wiring diagram of cellular function , volume =. Science , author =. 2016 , date =. doi:10.1126/science.aaf1420 , abstract =

    work page doi:10.1126/science.aaf1420 2016

  54. [88]

    Science , author =

    Systematic analysis of complex genetic interactions , volume =. Science , author =. 2018 , note =. doi:10.1126/science.aao1729 , abstract =

    work page doi:10.1126/science.aao1729 2018

  55. [89]

    Advances in Neural Information Processing Systems , publisher =

    Ke, Guolin and Meng, Qi and Finley, Thomas and Wang, Taifeng and Chen, Wei and Ma, Weidong and Ye, Qiwei and Liu, Tie-Yan , urldate =. Advances in Neural Information Processing Systems , publisher =. 2017 , file =

    2017

  56. [90]

    Not just a black box: Learning important features through propagating activation differences, 2017

    Shrikumar, Avanti and Greenside, Peyton and Shcherbina, Anna and Kundaje, Anshul , urldate =. Not Just a Black Box: Learning Important Features Through Propagating Activation Differences , url =. 2017 , langid =. doi:10.48550/arXiv.1605.01713 , shorttitle =. 1605.01713 [cs] , keywords =

    work page doi:10.48550/arxiv.1605.01713 2017

  57. [91]

    Journal of The Royal Society Interface , author =

    Testing the reproducibility and robustness of the cancer biology literature by robot , volume =. Journal of The Royal Society Interface , author =. 2022 , note =. doi:10.1098/rsif.2021.0821 , abstract =

    work page doi:10.1098/rsif.2021.0821 2022

  58. [92]

    Hierarchical deep learning for predicting

    Merino, Gabriela A and Saidi, Rabie and Milone, Diego H and Stegmayer, Georgina and Martin, Maria J , urldate =. Hierarchical deep learning for predicting. Bioinformatics , shortjournal =. 2022 , file =. doi:10.1093/bioinformatics/btac536 , abstract =

    work page doi:10.1093/bioinformatics/btac536 2022

  59. [93]

    Nature , author =

    Functional genomic hypothesis generation and experimentation by a robot scientist , volume =. Nature , author =. 2004 , langid =. doi:10.1038/nature02236 , abstract =

    work page doi:10.1038/nature02236 2004

  60. [94]

    Performance of Regression Models as a Function of Experiment Noise , volume =

    Li, Gang and Zrimec, Jan and Ji, Boyang and Geng, Jun and Larsbrink, Johan and Zelezniak, Aleksej and Nielsen, Jens and Engqvist, Martin. Performance of Regression Models as a Function of Experiment Noise , volume =. Bioinformatics and Biology Insights , shortjournal =. 2021 , langid =. doi:10.1177/11779322211020315 , abstract =

    work page doi:10.1177/11779322211020315 2021

  61. [95]

    Journal of The Royal Society Interface , author =

    Cheaper faster drug development validated by the repositioning of drugs against neglected tropical diseases , volume =. Journal of The Royal Society Interface , author =. 2015 , note =. doi:10.1098/rsif.2014.1289 , abstract =

    work page doi:10.1098/rsif.2014.1289 2015

  62. [96]

    Designing microplate layouts using artificial intelligence , volume =

    Francisco Rodríguez, María Andreína and Carreras Puigvert, Jordi and Spjuth, Ola , urldate =. Designing microplate layouts using artificial intelligence , volume =. Artificial Intelligence in the Life Sciences , shortjournal =. 2023 , keywords =. doi:10.1016/j.ailsci.2023.100073 , abstract =

    work page doi:10.1016/j.ailsci.2023.100073 2023

  63. [97]

    Genetics , shortjournal =

    Culbertson, Michael R and Henry, Susan A , urldate =. Genetics , shortjournal =. 1975 , file =. doi:10.1093/genetics/80.1.23 , abstract =

    work page doi:10.1093/genetics/80.1.23 1975

  64. [98]

    Yeast , author =

    Effect of benzoic acid on metabolic fluxes in yeasts: A continuous-culture study on the regulation of respiration and alcoholic fermentation , volume =. Yeast , author =. 1992 , langid =. doi:10.1002/yea.320080703 , shorttitle =

    work page doi:10.1002/yea.320080703 1992

  65. [99]

    2020 , eprint=

    Captum: A unified and generic model interpretability library for PyTorch , author=. 2020 , eprint=

    2020

  66. [100]

    doi:10.1093/gigascience/giad057 , shorttitle =

    Ma, Chunyu and Zhou, Zhihan and Liu, Han and Koslicki, David , urldate =. doi:10.1093/gigascience/giad057 , shorttitle =

    work page doi:10.1093/gigascience/giad057

  67. [101]

    and Jothi, J

    Syama, K. and Jothi, J. Angel Arul and Khanna, Namita , urldate =. Automatic disease prediction from human gut metagenomic data using boosting. doi:10.1186/s12859-023-05251-x , abstract =

    work page doi:10.1186/s12859-023-05251-x

  68. [102]

    Medical Entity Disambiguation Using Graph Neural Networks , isbn =

    Vretinaris, Alina and Lei, Chuan and Efthymiou, Vasilis and Qin, Xiao and Özcan, Fatma , urldate =. Medical Entity Disambiguation Using Graph Neural Networks , isbn =. Proceedings of the 2021 International Conference on Management of Data , publisher =. doi:10.1145/3448016.3457328 , series =

    work page doi:10.1145/3448016.3457328 2021

  69. [103]

    Dual Box Embeddings for the Description Logic

    Jackermeier, Mathias and Chen, Jiaoyan and Horrocks, Ian , urldate =. Dual Box Embeddings for the Description Logic. Proceedings of the. doi:10.1145/3589334.3645648 , series =

    work page doi:10.1145/3589334.3645648

  70. [104]

    International Conference on Principles of Knowledge Representation and Reasoning , author =

    From Knowledge Graph Embedding to Ontology Embedding?. International Conference on Principles of Knowledge Representation and Reasoning , author =

  71. [105]

    Wikidata: a free collaborative knowledgebase

    Vrandečić, Denny and Krötzsch, Markus , urldate =. Wikidata: a free collaborative knowledgebase , volume =. Communications of the ACM , abstract =. doi:10.1145/2629489 , shorttitle =

    work page doi:10.1145/2629489

  72. [106]

    Synergistic Signals: Exploiting Co-Engagement and Semantic Links via Graph Neural Networks , doi =

    Huang, Zijie and Li, Baolin and Asgharzadeh, Hafez and Cocos, Anne and Liu, Lingyi and Cox, Evan and Wise, Colby and Lamkhede, Sudarshan , date =. Synergistic Signals: Exploiting Co-Engagement and Semantic Links via Graph Neural Networks , doi =. 2312.04071 [cs] , keywords =

    work page arXiv

  73. [107]

    Proceedings of the 35th International Conference on Machine Learning , pages =

    A Semantic Loss Function for Deep Learning with Symbolic Knowledge , author =. Proceedings of the 35th International Conference on Machine Learning , pages =. 2018 , editor =

    2018

  74. [108]

    Contrastive Box Embedding for Collaborative Reasoning , isbn =

    Liang, Tingting and Zhang, Yuanqing and Di, Qianhui and Xia, Congying and Li, Youhuizi and Yin, Yuyu , urldate =. Contrastive Box Embedding for Collaborative Reasoning , isbn =. Proceedings of the 46th International. doi:10.1145/3539618.3591654 , series =

    work page doi:10.1145/3539618.3591654

  75. [109]

    When Box Meets Graph Neural Network in Tag-aware Recommendation , isbn =

    Lin, Fake and Zhao, Ziwei and Zhu, Xi and Zhang, Da and Shen, Shitian and Li, Xueying and Xu, Tong and Zhang, Suojuan and Chen, Enhong , urldate =. When Box Meets Graph Neural Network in Tag-aware Recommendation , isbn =. Proceedings of the 30th. doi:10.1145/3637528.3671973 , series =

    work page doi:10.1145/3637528.3671973

  76. [110]

    Representing Joint Hierarchies with Box Embeddings , doi =

    Dhruvesh Patel and Shib Sankar Dasgupta and Michael Boratko and Xiang Li and Luke Vilnis and Andrew McCallum , urldate =. Representing Joint Hierarchies with Box Embeddings , doi =. Automated Knowledge Base Construction (

  77. [111]

    Probabilistic Embedding of Knowledge Graphs with Box Lattice Measures , doi =

    Vilnis, Luke and Li, Xiang and Murty, Shikhar and. Probabilistic Embedding of Knowledge Graphs with Box Lattice Measures , doi =. Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , publisher =

  78. [112]

    and Naval, Prajakta and Bjurström, Erik Y

    Brunnsåker, Daniel and Gower, Alexander H. and Naval, Prajakta and Bjurström, Erik Y. and Kronström, Filip and Tiukova, Ievgeniia A. and King, Ross D. , urldate =. Agentic. doi:10.1101/2025.06.24.661378 , shorttitle =

    work page doi:10.1101/2025.06.24.661378 2025

  79. [113]

    19th International Conference on Neurosymbolic Learning and Reasoning , year=

    Ontology-based box embeddings and knowledge graphs for predicting phenotypic traits in Saccharomyces cerevisiae , author=. 19th International Conference on Neurosymbolic Learning and Reasoning , year=

  80. [114]

    Poincaré embeddings for learning hierarchical representations , isbn =

    Nickel, Maximilian and Kiela, Douwe , date =. Poincaré embeddings for learning hierarchical representations , isbn =. 2017 , booktitle =

    2017

Showing first 80 references.