Correlations from structure and phylogeny combine constructively in the inference of protein partners from sequences

Abstract

Inferring protein-protein interactions from sequences is an important task in computational biology. Recent methods based on Direct Coupling Analysis (DCA) or Mutual Information (MI) allow to find interaction partners among paralogs of two protein families. Does successful inference mainly rely on correlations from structural contacts or from phylogeny, or both? Do these two types of signal combine constructively or hinder each other? To address these questions, we generate and analyze synthetic data produced using a minimal model that allows us to control the amounts of structural constraints and phylogeny. We show that correlations from these two sources combine constructively to increase the performance of partner inference by DCA or MI. Furthermore, signal from phylogeny can rescue partner inference when signal from contacts becomes less informative, including in the realistic case where inter-protein contacts are restricted to a small subset of sites. We also demonstrate that DCA-inferred couplings between non-contact pairs of sites improve partner inference in the presence of strong phylogeny, while deteriorating it otherwise. Moreover, restricting to non-contact pairs of sites preserves inference performance in the presence of strong phylogeny. In a natural dataset, as well as in realistic synthetic data based on it, we find that non-contact pairs of sites contribute positively to partner inference performance, and that restricting to them preserves performance, evidencing an important role of phylogeny.Author summaryIn protein sequence data, the amino acid usages at different sites of a protein or of two interacting proteins can be correlated because of functional constraints. For instance, the need to maintain physico-chemical complementarity among two sites that are in contact in the three-dimensional structure of a protein complex causes such correlations. However, correlations can also arise due to shared evolutionary history, even in the absence of any functional constraint. While these phylogenetic correlations are known to obscure the inference of structural contacts, we show, using controlled synthetic data, that correlations from structure and phylogeny combine constructively to allow the inference of protein partners from sequences. We also show that pairs of amino acids that are not in contact in the structure have a major impact on partner inference in a natural dataset and in realistic synthetic ones. These findings explain the success of methods based on pairwise maximum-entropy models or on information theory at predicting protein partners from sequences.

Impact of phylogeny on structural contact inference from protein sequence data

“…Note that, in Ref. [32], we showed that these pairs of sites are particularly useful for the inference of interacting partners among the paralogs of two protein families, where phylogenetic correlations are useful.…”

Combining phylogeny and coevolution improves the inference of interaction partners among paralogous proteins

“…While unifying these two signals in a single framework remains a hard problem, they have been shown to combine constructively in the inference of protein partners among paralogs by coevolution methods [24], raising the possibility that explicitly combining them may further increase performance. Here, we combine these two signals in a technically straightforward way.…”

“…Indeed, the limited numbers of paralogs per species in these datasets make the IPA already very efficient and robust without any seed co-MSA. Note that the successful performance of IPA on pairs of protein families with no known interaction shows that it is already able to exploit phylogenetic signal [23,24]. With GA, this signal is exploited more explicitly.…”

“…However, while these methods can function even without an initial seed co-MSA, their performance remains limited for high paralog numbers as well as for small datasets [21]. Interestingly, these coevolution-based methods already benefit from phylogenetic correlations [23,24], and mutual information performs slightly better than DCA for the pairing task [25], already hinting at the potential of combining the two signals. However, coevolution-based methods are unlikely to exploit sequence similarity in an optimal way, because they are not designed for that.…”

Generative power of a protein language model trained on multiple sequence alignments

“…However, as bmDCA and MSA Transformer are trained on natural data, they can capture phylogenetic correlations [53]. Besides, inferred Potts models are known to be impacted by phylogenetic correlations [8,10,[65][66][67][68][69][70][71].…”