henhao
Sign in
BiophysicsbioRxiv
Preprint

Functional pocketomes and complete allosteric maps of five protein kinases

Carla FolgadoiD, Toni BeltraniD, Ben LehneriD

PDFDOICode
Post to X
Threads on XPost to BlueskyThreads on Bluesky
In-depth full-text digest available · Jump to digest ↓

Abstract

Protein kinases are central to regulation, dysregulated in many diseases, and the targets of a hundred clinically-approved drugs. Structural conservation of kinase active sites makes the development of selective inhibitors challenging. Targeting allosteric sites can increase selectivity, overcome resistance mutations, and activate kinases. However, most kinases have no known allosteric sites. Moreover, the conservation of allosteric regulation in proteins with the same structural fold is unclear. Here, we address this question by using >390,000 measurements of mutational effects on activity and abundance to construct complete comparative energetic and allosteric maps of five human protein kinase domains: SRC, FGR, JNK2/MAPK9, ZAK/MAP3K20, and TSSK2. For inhibition, all five kinases have distance-dependent allosteric regulation. However allostery is anisotropic, and each protein has a distinct allosteric architecture and surface to therapeutically target. A set of inhibitory allosteric pockets close to the active site is conserved in all five proteins, but more distal allosteric pockets are protein-specific. These include pockets that are activatory in one kinase and inhibitory in another. Differences in allosteric regulation are particularly striking for activation, where the allosteric maps are highly diverged. Our results show that proteins with the same structural fold have different energetic architectures and allosteric pockets. Mapping the 'functional pocketomes' of kinases and other clinically-important targets should empower the development of more selective therapeutics.

02 /Specification & Classification

Click any tag to browse matching papers
Citations
2
OpenAlex Topics

In-depth summary

✦ Multi-agent verified full-text synthesis

Claude Sonnet 4.6 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (2 Nov 2026) · CC BY 4.0

Energy-model maps of five human kinases show conserved distance-dependent decay of inhibitory allostery, but distal allosteric pockets and activating sites diverge between kinases.

Most kinases have no known allosteric sites, the conserved ATP pocket makes selective inhibitors hard to develop, and it is unclear whether proteins sharing a fold share allosteric regulation. This work builds directly comparable experimental allosteric maps for five kinases to test how much energetic architecture and functional pockets are conserved.

Key results

  1. ShownInhibitory allostery decays with distance from the active site in all five kinases, though SRC decays sharpest. Distance explains only a minority of the variance in ΔΔGa for inactivating mutations.

    Median inhibitory-effect half-distance d½ 9.3 Å across five proteins, from 6.5 Å (SRC) to 11.1 Å (ZAK, 95% CI 9.0-14.3). For inactivating mutations, distance explains on average 29% of ΔΔGa variance (23% ZAK to 36% JNK2). · Conserved distance-dependent allosteric decay…; Anisotropic allostery

  2. ShownAllosteric coupling is anisotropic: strongest toward helix αC and weakest toward helix αD. The strength in a given direction differs between kinases.

    Median d½ by direction: 26.1 Å (y+, αC), 18.6 Å (z+), 14.1 Å (z-), 4.8 Å (x+). Toward the C-lobe, FGR 28.6 Å vs SRC 15.2 Å (z = 6.3, p < 0.05). Toward the activation loop, JNK2 46.4 Å vs ZAK 14.9 Å (z = 5.3, p < 0.05). · Anisotropic allostery

  3. ShownA cluster of inhibitory pockets in the interlobe cleft next to the active site is allosteric in all five kinases. These pockets are less structurally conserved than the ATP site but more conserved than the average pocket.

    FTMap found 37 pockets; 30 were enriched for allosteric mutations in at least one kinase. Five were inhibitory in all five: ATP (10, 11), DFG, MT3, plus uncatalogued 19 and 20. Cα RMSD was 1.82 (DFG), 0.92 (MT3), 2.03 and 1.93 Å vs 0.74 and 0.77 Å for the ATP pockets. · A cluster of allosteric pockets conserved in all…

  4. ShownActivating mutations are dispersed across the domain, not enriched near the active site, and are mostly specific to a single kinase. Activation pockets are less conserved than inhibitory ones.

    43 of 47 activating-enriched residues are enriched in one protein only. 860 of 946 distinct activating mutations act in a single kinase and none in all five. 13 of 17 activating pockets are single-kinase, vs 9 of 22 inhibitory pockets. · Activating mutations are structurally dispersed…; Activation pockets

  5. ShownPockets can switch sign between kinases, and even the close homologs SRC and FGR differ in functional pockets, suggesting routes to selective modulators.

    Seven pockets activate in one kinase and inhibit in another; pocket 2 is activating in TSSK2 (OR 26.9) but inhibitory in SRC, FGR, ZAK (OR 1.5-2.7). Pockets 1, 22, 28, 35 are inhibitory in SRC (OR 2.7, 1.9, 2.3, 4.1) but not FGR (OR 0.9, 0.8, 1.0, 1.2). · Selectively targeting highly-related kinases; Pockets that switch from activation to…

Methods, data and reproducibility

KINASE-MAPS: double-mutant libraries of each kinase domain were selected in yeast for activity-dependent toxicity and folded abundance (sPCA). A three-state thermodynamic model (MoCHI) inferred ΔΔGf and ΔΔGa per substitution, and AlphaFold3 models with FTMap defined distances and pockets.

  • Double-mutant kinase libraries, activity and abundance selections in yeast; FGR, JNK2, ZAK, TSSK2 new. Design over five kinases: 396,800 nt variants, 248,000 aa sequences · human kinases in Saccharomyces cerevisiae · n = 163,185 activity and 231,716 abundance aa variants (five kinases) · ENA PRJEB122867
  • SRC data previously published, included in the five-kinase comparison · human
  • Inferred free-energy changes across the five kinases · human · n = 50,396 (26,032 ΔΔGf; 24,711 ΔΔGa)
  • AlphaFold3 ATP-bound models of the five kinase domains, used for distances, SASA, contacts and pocket mapping · human · n = 5 models
  • FTMap fragment-sized surface pockets compared with the Kinase Atlas · human · n = 37 pockets
Methods & tools
MoCHI, DiMSum, SUNi mutagenesis, sPCA (DHFR protein-fragment complementation), activity-dependent toxicity assay, AlphaFold3, FTMap, Kinase Atlas, KinCore alignment, freeSASA v2.2.1, getContacts, alignparse
Code
https://github.com/lehner-lab/KINASE-MAPS
Software & environment
freeSASA v2.2.1; DiMSum run with default parameters plus two adjustments; MoCHI for model fitting; AlphaFold3 web server; exponential fits with nls in R.
Reproducibility gaps
Sequencing data are given only as ENA accession PRJEB122867 (no URL printed); no container or compute details are stated; SRC data were previously published.
Limitations (3)
  • The authors note that the mechanistic reasons why mutational effects on fold stability diverge between kinases remain to be understood, and that this has been little studied in larger domains. · stated by the authors
  • Activity is measured as yeast toxicity of overexpressed full-length human kinases, and the energy model assumes additive, phenomenological three-state behaviour; effects in native cells or on physiological substrates are not tested. · noted in this summary
  • Pockets are defined computationally (FTMap on AlphaFold3 models) and linked to mutational enrichment only; no ligand binding or compound-induced modulation was tested, and only five kinases were sampled. · noted in this summary

Who should read:Read this if you work on allosteric drug discovery or deep mutational scanning and want comparative maps of conserved versus kinase-specific functional pockets.

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISBeltran '26Godbole '23Chauhan '26Imamoto '26Ma '26Al-Masri '22
76% semantic matchBiophysics
Toni Beltran, Mohsin M. Naqvi, André J. Faure+1 more·Science Advances·Feb 11, 2026· 25 citations