Machine Learnable Fold Space Representation based on Residue Cluster Classes

Corral-Corral, Ricardo; Chávez, Edgar; Río, Gabriel del

doi:10.1016/j.compbiolchem.2015.07.010

Cited by 13 publications

(12 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We previously noted that the size (total number of amino acid residues) of proteins is linearly related to the number of clusters in the RCC [21], consistent with experimental observations that proteins present a near-constant density [37]. Based on our current results indicating that adding RCCs effectively represent PPI, it seems that protein-protein complexes may also linearly grow in number of clusters according to the number of residues (the sum of RCCs linearly increases the number of clusters with the molecular weight of the protein complex), or simply keep a constant density.…”

Section: Discussionmentioning

confidence: 99%

“…We showed that RCCs present a pattern that is recognizable by any heuristic ML approach and consequently provided a learnable representation for protein structure classification. Indeed, we showed that RCCs improved upon the state-of-the-art methods aimed to look for structural neighbors and structural classification [21]. In the present work, we aimed to test if protein structure is key to protein interactions through the use of RCCs to classify PPI.…”

Section: Of 17mentioning

confidence: 91%

“…In the present work, we aimed to test if protein structure is key to protein interactions through the use of RCCs to classify PPI. We previously reported a compact, 26-dimensional representation of protein structure based on residue cluster classes (RCCs) [21], which are obtained from the maximal cliques observed in the contact map derived from the protein 3D structure, taking into account the sequence localization of amino acid residues. RCCs are sets of maximal cliques that are grouped by size and classified according to the sequence proximity of the included residues (see Figure 1).…”

Section: Of 17mentioning

confidence: 99%

See 2 more Smart Citations

Protein–Protein Interactions Efficiently Modeled by Residue Cluster Classes

Velez

Fontove²,

Río

2020

IJMS

Self Cite

View full text Add to dashboard Cite

Predicting protein–protein interactions (PPI) represents an important challenge in structural bioinformatics. Current computational methods display different degrees of accuracy when predicting these interactions. Different factors were proposed to help improve these predictions, including choosing the proper descriptors of proteins to represent these interactions, among others. In the current work, we provide a representative protein structure that is amenable to PPI classification using machine learning approaches, referred to as residue cluster classes. Through sampling and optimization, we identified the best algorithm–parameter pair to classify PPI from more than 360 different training sets. We tested these classifiers against PPI datasets that were not included in the training set but shared sequence similarity with proteins in the training set to reproduce the situation of most proteins sharing sequence similarity with others. We identified a model with almost no PPI error (96–99% of correctly classified instances) and showed that residue cluster classes of protein pairs displayed a distinct pattern between positive and negative protein interactions. Our results indicated that residue cluster classes are structural features relevant to model PPI and provide a novel tool to mathematically model the protein structure/function relationship.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Of 17mentioning

confidence: 91%

Section: Of 17mentioning

confidence: 99%

See 1 more Smart Citation

Protein–Protein Interactions Efficiently Modeled by Residue Cluster Classes

Velez

Fontove²,

Río

2020

IJMS

Self Cite

View full text Add to dashboard Cite

show abstract

“…We built different RCC representations of protein structures to identify the best distance criterion to classify 3D protein structures. These included distances at 5, 6, 7, 8, 9, 10 and 15 Å; we originally used only a distance of 5 Å [11]; hence, this exercise allowed us to compare the efficiency of RCCs previously reported. In addition, we also included a variant in the construction of RCCs: the inclusion or exclusion of the amino acid side-chain atoms.…”

Section: Protein Structural Classificationmentioning

confidence: 99%

“…We have previously described a representation of 3D structure that is learnable (that is, it displays a pattern that any ML heuristic model should be able to detect); such a representation allowed us to identify structural neighbors and classify the protein's 3D structure with the best performance and reliability reported so far [11]. The representation is based on counting the 26 different maximal clique classes that are derived from the 3D structure and protein sequence given a contact distance threshold of 5 Å, including atoms of the side chains; we referred to these maximal cliques as residue cluster classes or RCCs (see Figure 1 and Materials and Methods).…”

Section: Introductionmentioning

confidence: 99%

Residue Cluster Classes: A Unified Protein Representation for Efficient Structural and Functional Classification

Fontove¹,

Río

2020

Entropy

Self Cite

View full text Add to dashboard Cite

Proteins are characterized by their structures and functions, and these two fundamental aspects of proteins are assumed to be related. To model such a relationship, a single representation to model both protein structure and function would be convenient, yet so far, the most effective models for protein structure or function classification do not rely on the same protein representation. Here we provide a computationally efficient implementation for large datasets to calculate residue cluster classes (RCCs) from protein three-dimensional structures and show that such representations enable a random forest algorithm to effectively learn the structural and functional classifications of proteins, according to the CATH and Gene Ontology criteria, respectively. RCCs are derived from residue contact maps built from different distance criteria, and we show that 7 or 8 Å with or without amino acid side-chain atoms rendered the best classification models. The potential use of a unified representation of proteins is discussed and possible future areas for improvement and exploration are presented.

show abstract