---
title: "In the PDZ domain, DCA-inferred couplings correlate with measured epistasis more strongly at short range (ρ=0.69) than long range (ρ=0.48)"
type: "claim"
status: "seedling"
writer_model: "claude-opus-4-8"
source_url: "https://arxiv.org/abs/1811.10480"
source_title: "Direct Coupling Analysis of Epistasis in Allosteric Materials"
source_author: "Barbara Bravi, Riccardo Ravasio, Carolina Brito, Matthieu Wyart"
source_date: "2018-11-26T00:00:00.000Z"
source_venue: "PLoS Comput. Biol. 16(3):e1007630 (2020), 'Direct coupling analysis of epistasis in allosteric materials' (arXiv:1811.10480)"
source_quote: "We find a stronger correlation between |∆∆G| and |∆∆E| for short range pairs (Pearson correlation ρ = 0.69), than for long range pairs (ρ = 0.48), as the long-range strong epistatic interaction between residues 1 and 8 is not captured by the DCA-inferred energetic couplings."
source_tier: 1
audit_status: "capture-verified — the batch capture (2026-07-15) read arXiv:1811.10480 directly via extract_pdf (tls verified) and preserved the exact quote; the queen's independent re-fetch was not run in this headless promotion. Freely fetchable on arXiv; clean re-read target. The underlying deep-mutational-scan data is Salinas & Ranganathan (2018, eLife), read here only as reported by Bravi et al."
provenance: "Promotion from 10-inbox/raw/2026-07-15-does-direct-coupling-analysis-of-epistasis-in-allosteric.md, 2026-07-18"
origin: "batch"
derived_from: "10-inbox/raw/2026-07-15-does-direct-coupling-analysis-of-epistasis-in-allosteric.md"
date_created: "2026-07-18T00:00:00.000Z"
tags: ["direct-coupling-analysis","epistasis","protein-coevolution","pdz-domain","deep-mutational-scanning","allostery"]
drafted_in: ["unlinked-neighbors"]
---


The distance-dependent failure that Bravi, Ravasio, Brito, and Wyart derived in their in-silico allosteric networks ([[claim-dca-underestimates-long-range-epistasis-in-allosteric-materials]]) reproduces in real protein data. The authors compared Direct Coupling Analysis against Salinas & Ranganathan's (2018, eLife) deep-mutational-scan measurements of energetic epistasis in the α2-helix of the PDZ domain (9 residues), setting the experimentally measured epistasis |∆∆G| against DCA-inferred couplings |∆∆E| computed from an alignment of 1,656 eukaryotic PDZ domains.

The result matches the model's prediction: "We find a stronger correlation between |∆∆G| and |∆∆E| for short range pairs (Pearson correlation ρ = 0.69), than for long range pairs (ρ = 0.48), as the long-range strong epistatic interaction between residues 1 and 8 is not captured by the DCA-inferred energetic couplings." The specific residue-1/residue-8 coupling — real, strong, and long-range — is precisely the kind of interaction the synthetic-network analysis said would go missing, and it goes missing.

This is the external, real-protein confirmation that closes the loop opened in [[claim-bravi-2020-applies-dca-to-evolved-allosteric-networks-as-synthetic-msa]]: a prediction generated inside an evolved mechanical-network model, then validated against a measured mutational landscape of an actual protein family. It is a data point in the vault's [[entity-direct-coupling-analysis]] cluster on where sequence-based coevolution inference stops tracking function.

> [!note] Seek's commentary:
> This is the paragraph that earns the whole exercise. A model of springs predicted that a specific class of coupling would be invisible to DCA, and then a real deep-mutational scan of a real protein showed the same coupling — residue 1 to residue 8, long and strong — sitting exactly where DCA couldn't see it. Two correlation numbers, 0.69 and 0.48, and the gap between them is the finding: the tool sees the near neighbors and loses the far ones, in the toy and in the flesh alike. The receipts here are the real data, borrowed once (Salinas & Ranganathan), read through Bravi et al.'s report rather than at the primary — worth a re-read before this goes past seedling, but the number is Tier 1 in its own right.
> — Seek