FingerprintContacts: Predicting Alternative Conformations of Proteins from Coevolution

Abstract

<jats:title>Abstract</jats:title><jats:p>Proteins are dynamic molecules which perform diverse molecular functions by adopting different three-dimensional structures. Recent progress in residue-residue contacts prediction opens up new avenues for the <jats:italic>de novo</jats:italic> protein structure prediction from sequence information. However, it is still difficult to predict more than one conformation from residue-residue contacts alone. This is due to the inability to deconvolve the complex signals of residue-residue contacts, i.e. spatial contacts relevant for protein folding, conformational diversity, and ligand binding. Here, we introduce a machine learning based method, called FingerprintContacts, for extending the capabilities of residue-residue contacts. This algorithm leverages the features of residue-residue contacts, that is, (1) a single conformation outperforms the others in the structural prediction using all the top ranking residue-residue contacts as structural constraints, and (2) conformation specific contacts rank lower and constitute a small fraction of residue-residue contacts. We demonstrate the capabilities of FingerprintContacts on eight ligand binding proteins with varying conformational motions. Furthermore, FingerprintContacts identifies small clusters of residue-residue contacts which are preferentially located in the dynamically fluctuating regions. With the rapid growth in protein sequence information, we expect FingerprintContacts to be a powerful first step in structural understanding of protein functional mechanisms.</jats:p>

Exploring the alternative conformation of a known protein structure based on contact map prediction

“…31,32,44 Here we further expand the idea to deep-learning-based CM prediction and provide strong evidence for de novo prediction of multistate structural information by using deep learning, in addition to Feng and Shukla’s work that utilized a different strategy. 38…”

“…31,32,44 Here we further expand the idea to deep-learning-based CM prediction and provide strong evidence for de novo prediction of multistate structural information by using deep learning, in addition to Feng and Shukla's work that utilized a different strategy. 38 Figure 5: Analysis of contacts in the predicted contact maps (PCMs) for the additional 91 proteins. The number of apo/holo-state speciic contacts ("101"/"011" in table S3) were represented on the x/y axis.…”

“…Based on pioneering DCA studies, deep learning-based prediction of protein contact maps (CMs) has been improved and widely used for 3D protein structure predictions in the last few years, 6,35,36 but the biological importance of CM prediction is not as immediately evident as that of DCA. Notably, Iyer et al found that the difference of CMs from PDB structures can infer the conformational flexibility of proteins, 37 and Feng et al developed a method to predict the alternative conformations of proteins by using contacts clustering and Confold2, 38 strongly suggesting that CM prediction may contain multistate structural information. However, neither the aforementioned AlphaFold-based nor CM-based studies excluded the homologous proteins from the training dataset.…”

HOPMA: Boosting protein functional dynamics with colored contact maps

“…This set up allows obtaining the optimal (or close-to-optimal) transitions within our framework. In practical applications, when the target is not know, the sampling may be guided by additional information (83, 84), such as experimental small-angle scattering profiles (25–27, 85, 86), low-resolution electron density maps (15–22), intensities from crystallographic scattering experiments (87, 88), complementarity of the molecular shapes upon binding (28–36), or other type of data such as contacts inferred from coevolutionary signals extracted from protein sequences (89–91).…”