Anc2vec: embedding gene ontology terms by preserving ancestors relationships

Abstract

The gene ontology (GO) provides a hierarchical structure with a controlled vocabulary composed of terms describing functions and localization of gene products. Recent works propose vector representations, also known as embeddings, of GO terms that capture meaningful information about them. Significant performance improvements have been observed when these representations are used on diverse downstream tasks, such as the measurement of semantic similarity between GO terms and functional similarity between proteins. Despite the success shown by these approaches, existing embeddings of GO terms still fail to capture crucial structural features of the GO. Here, we present anc2vec, a novel protocol based on neural networks for constructing vector representations of GO terms by preserving three important ontological features: its ontological uniqueness, ancestors hierarchy and sub-ontology membership. The advantages of using anc2vec are demonstrated by systematic experiments on diverse tasks: visualization, sub-ontology prediction, inference of structurally related terms, retrieval of terms from aggregated embeddings, and prediction of protein–protein interactions. In these tasks, experimental results show that the performance of anc2vec representations is better than those of recent approaches. This demonstrates that higher performances on diverse tasks can be achieved by embeddings when the structure of the GO is better represented. Full source code and data are available at https://github.com/sinc-lab/anc2vec.

E-SNPs&GO: Embedding of protein sequence and function improves the annotation of human pathogenic variants

“…In particular, several embedding models, have been defined to provide a numerical representation of nodes in ontologies (Zhong et al ., 2020; Chen et al ., 2021). Here we adopt Anc2Vec (Edera et al ., 2022), a method that learns a vector representation for GO terms, by preserving ancestors relationships.…”

“…Terms from the three different GO sub-ontologies (Molecular Function, Cellular Component, Biological Process) are processed independently. Each annotated GO term is then embedded as a vector of 200 features using the Anc2Vec model (Edera et al ., 2022). To obtain a single vector representation independent of the number of terms of a given protein, we average all the vector encodings (Figure 1).…”

“…In the input encoding phase, the sequence and its variant are embedded with two different procedures, ESM-1v (Meier et al, 2021) and ProtTrans T5 (Elnaggar et al, 2021). In order to embed the functional protein annotation of the wildtype protein, we adopt Anc2Vec (Edera et al, 2022).…”

“…Firstly, we adopt two different and complementary embedding procedures, ProtTrans T5 (Elnaggar et al, 2021) and ESM-1v (Meier et al, 2021), to directly encode an input variation without introducing any hand-crafted feature as previously done. Secondly, leveraging the idea introduced in SNPs&GO (Calabrese et al, 2009), we explore a new way of encoding functional annotations by adopting a model called Anc2Vec (Edera et al, 2022), specifically designed for the embedding of Gene Ontology (GO) terms.…”

“…Not limited to the encoding of protein sequence data, embedding techniques are also applied to model the relationships existing within more complex structures such as graphs, networks, or biological ontologies (Grover and Leskovec, 2016; Perozzi et al ., 2014; Zhong et al ., 2020; Kandathil et al ., 2022; Edera et al ., 2022).…”