Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

INDIGENA: inductive prediction of disease-gene associations using phenotype ontologies

Fernando Zhapa-Camacho, Robert Hoehndorf

TL;DR

INDIGENA addresses the challenge of predicting disease–gene associations for Mendelian diseases by leveraging phenotype ontologies in an inductive framework. It constructs cross-species phenotype graphs, learns latent phenotype representations with multiple graph embedding models, and aggregates these embeddings to score gene–disease pairs, incorporating a supervised signal from known GDAs. The approach demonstrates that inductive, phenotype-based embeddings can approach transductive performance while outperforming handcrafted semantic similarity baselines, with strong statistical support. This has practical implications for rare-disease diagnosis and clinical genomics, enabling gene prioritization for novel diseases described by phenotypes without retraining on those diseases.

Abstract

Motivation: Predicting gene-disease associations (GDAs) is the problem to determine which gene is associated with a disease. GDA prediction can be framed as a ranking problem where genes are ranked for a query disease, based on features such as phenotypic similarity. By describing phenotypes using phenotype ontologies, ontology-based semantic similarity measures can be used. However, traditional semantic similarity measures use only the ontology taxonomy. Recent methods based on ontology embeddings compare phenotypes in latent space; these methods can use all ontology axioms as well as a supervised signal, but are inherently transductive, i.e., query entities must already be known at the time of learning embeddings, and therefore these methods do not generalize to novel diseases (sets of phenotypes) at inference time. Results: We developed INDIGENA, an inductive disease-gene association method for ranking genes based on a set of phenotypes. Our method first uses a graph projection to map axioms from phenotype ontologies to a graph structure, and then uses graph embeddings to create latent representations of phenotypes. We use an explicit aggregation strategy to combine phenotype embeddings into representations of genes or diseases, allowing us to generalize to novel sets of phenotypes. We also develop a method to make the phenotype embeddings and the similarity measure task-specific by including a supervised signal from known gene-disease associations. We apply our method to mouse models of human disease and demonstrate that we can significantly improve over the inductive semantic similarity baseline measures, and reach a performance similar to transductive methods for predicting gene-disease associations while being more general. Availability and Implementation: https://github.com/bio-ontology-research-group/indigena

INDIGENA: inductive prediction of disease-gene associations using phenotype ontologies

TL;DR

INDIGENA addresses the challenge of predicting disease–gene associations for Mendelian diseases by leveraging phenotype ontologies in an inductive framework. It constructs cross-species phenotype graphs, learns latent phenotype representations with multiple graph embedding models, and aggregates these embeddings to score gene–disease pairs, incorporating a supervised signal from known GDAs. The approach demonstrates that inductive, phenotype-based embeddings can approach transductive performance while outperforming handcrafted semantic similarity baselines, with strong statistical support. This has practical implications for rare-disease diagnosis and clinical genomics, enabling gene prioritization for novel diseases described by phenotypes without retraining on those diseases.

Abstract

Motivation: Predicting gene-disease associations (GDAs) is the problem to determine which gene is associated with a disease. GDA prediction can be framed as a ranking problem where genes are ranked for a query disease, based on features such as phenotypic similarity. By describing phenotypes using phenotype ontologies, ontology-based semantic similarity measures can be used. However, traditional semantic similarity measures use only the ontology taxonomy. Recent methods based on ontology embeddings compare phenotypes in latent space; these methods can use all ontology axioms as well as a supervised signal, but are inherently transductive, i.e., query entities must already be known at the time of learning embeddings, and therefore these methods do not generalize to novel diseases (sets of phenotypes) at inference time. Results: We developed INDIGENA, an inductive disease-gene association method for ranking genes based on a set of phenotypes. Our method first uses a graph projection to map axioms from phenotype ontologies to a graph structure, and then uses graph embeddings to create latent representations of phenotypes. We use an explicit aggregation strategy to combine phenotype embeddings into representations of genes or diseases, allowing us to generalize to novel sets of phenotypes. We also develop a method to make the phenotype embeddings and the similarity measure task-specific by including a supervised signal from known gene-disease associations. We apply our method to mouse models of human disease and demonstrate that we can significantly improve over the inductive semantic similarity baseline measures, and reach a performance similar to transductive methods for predicting gene-disease associations while being more general. Availability and Implementation: https://github.com/bio-ontology-research-group/indigena
Paper Structure (15 sections, 9 equations, 2 figures, 3 tables)

This paper contains 15 sections, 9 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Materials and Methods
  3. Datasets
  4. Ontology preprocessing and graph projection
  5. Graph Embedding Methods
  6. Semantic Similarity
  7. Embedding-based similarity
  8. Results
  9. Transductive prediction of gene--disease associations
  10. Inductive Approach for Gene--Disease Association Prediction
  11. Discussion
  12. Inductive Gene--Disease Association Prediction
  13. Embedding model performance
  14. Comparison with Semantic Similarity Measures
  15. Conclusions

Figures (2)

  • Figure 1: Different graph structures. Nodes $\bullet$ represent UPheno entities, $\bullet$ represent genes, $\bullet$ represent training diseases and $\bullet$ represent testing diseases. (a) Graph 1 is the original UPheno graph representation. (b) Graph 2 includes gene--phenotype associations. (c) Graph 3 includes disease--phenotype associations. (d) Graph 4 includes gene--disease associations. (e) and (f) are transductive variations where testing diseases have been added to the graph linking them to their phenotypes.
  • Figure 2: UMAP representation of learned embeddings for methods TransE, TransH and TransD on the first fold. TransE learns a distance function direcly, therefore, their embeddings are scattered across the latent space. TransH and TransD learn a projection function; therefore the initial latent space capture similarity features of embeddings.