Poincaré maps for visualization of large protein families

Susmelj, Anna Klimovskaia; Ren, Yani; Meersche, Yann Vander; Gelly, Jean‐Christophe; Galochkina, Tatiana

doi:10.1093/bib/bbad103

“…Utilizing Poincaré maps to embed kinase hierarchies is not a novel concept. In a related method, PoincaréMSA [ 44 ], kinase proteins were embedded in an unsupervised way. However, in our approach, instead of multiple sequence alignment data, we utilized ProtVec as input.…”

Section: Resultsmentioning

confidence: 99%

“…The early use cases of manual Poincaré embeddings for cluster identification and hierarchy visualization [ 43 ] have evolved with automatic learning methods. For example, the PoincaréMSA method automatically produces kinase hierarchy embeddings with the Poincaré maps dimensionality reduction technique using multiple sequence alignment data [ 44 ].…”

Section: Preliminariesmentioning

confidence: 99%

Towards explainable interaction prediction: Embedding biological hierarchies into hyperbolic interaction space

Pogány,

Antal

2024

PLoS ONE

0

View full text Add to dashboard Cite

Given the prolonged timelines and high costs associated with traditional approaches, accelerating drug development is crucial. Computational methods, particularly drug-target interaction prediction, have emerged as efficient tools, yet the explainability of machine learning models remains a challenge. Our work aims to provide more interpretable interaction prediction models using similarity-based prediction in a latent space aligned to biological hierarchies. We investigated integrating drug and protein hierarchies into a joint-embedding drug-target latent space via embedding regularization by conducting a comparative analysis between models employing traditional flat Euclidean vector spaces and those utilizing hyperbolic embeddings. Besides, we provided a latent space analysis as an example to show how we can gain visual insights into the trained model with the help of dimensionality reduction. Our results demonstrate that hierarchy regularization improves interpretability without compromising predictive performance. Furthermore, integrating hyperbolic embeddings, coupled with regularization, enhances the quality of the embedded hierarchy trees. Our approach enables a more informed and insightful application of interaction prediction models in drug discovery by constructing an interpretable hyperbolic latent space, simultaneously incorporating drug and target hierarchies and pairing them with available interaction information. Moreover, compatible with pairwise methods, the approach allows for additional transparency through existing explainable AI solutions.

show abstract

“…The early use cases of manual Poincaré embeddings for cluster identification and hierarchy visualization [33] have evolved with automatic learning methods. For example, the PoincaréMSA method automatically produces kinase hierarchy embeddings with the Poincaré maps dimensionality reduction technique using multiple sequence alignment data [34].…”

Section: Preliminariesmentioning

confidence: 99%

“…Utilizing Poincaré maps to embed kinase hierarchies is not a novel concept. In a related method, PoincaréMSA [34], kinase proteins were embedded in an unsupervised way. However, in our approach, instead of multiple sequence alignment data, we utilized ProtVec as input.…”

Section: Latent Space Analysismentioning

confidence: 99%

Towards explainable interaction prediction: Embedding biological hierarchies into hyperbolic interaction space

Pogány,

Antal

2023

Preprint

0

View full text Add to dashboard Cite

Given the prolonged timelines and high costs associated with traditional approaches, accelerating drug development is crucial. Computational methods, particularly drug-target interaction prediction, have emerged as efficient tools, yet the explainability of machine learning models remains a challenge. Our work aims to provide more interpretable interaction prediction models using similarity-based prediction in a latent space aligned to biological hierarchies. We investigated integrating drug and protein hierarchies into a joint-embedding drug-target latent space via embedding regularization by conducting a comparative analysis between models employing traditional flat Euclidean vector spaces and those utilizing hyperbolic embeddings. Besides, we provided a latent space analysis as an example to show how we can gain visual insights into the trained model with the help of dimensionality reduction. Our results demonstrate that hierarchy regularization improves interpretability without compromising predictive performance. Furthermore, integrating hyperbolic embeddings, coupled with regularization, enhances the quality of the embedded hierarchy trees. Our approach enables a more informed and insightful application of interaction prediction models in drug discovery by constructing an interpretable hyperbolic latent space, simultaneously incorporating drug and target hierarchies and pairing them with available interaction information. Moreover, compatible with pairwise methods, the approach allows for additional transparency through existing explainable AI solutions.

show abstract

MMV method: a new approach to compare protein sequences under binary representation

Pal,

Ghosh,

Maji

et al. 2024

Journal of Biomolecular Structure and Dynamics

1

0

View full text Add to dashboard Cite

Poincaré maps for visualization of large protein families

Cited by 3 publications

References 25 publications

Towards explainable interaction prediction: Embedding biological hierarchies into hyperbolic interaction space

Towards explainable interaction prediction: Embedding biological hierarchies into hyperbolic interaction space

Towards explainable interaction prediction: Embedding biological hierarchies into hyperbolic interaction space

MMV method: a new approach to compare protein sequences under binary representation

Contact Info

Product

Resources

About