Allostery: Theoretical framework of conformational behavior

Over three decades ago, a landmark perspective laid down a foundational view of the free energy landscape of proteins (Frauenfelder et al., Reference Frauenfelder1991), which led to the transformed concept of molecular recognition and allostery (Nussinov et al., Reference Nussinov2019a). The prevailing view, which derived largely from crystallography, has been of proteins in their inactive – or active – structures. The Frauenfelder, Sligar, and Wolynes’ highly influential portrayal of proteins as an ensemble of a vast number of states, plotted in two-dimensional graphs by their free energies, offered a new way of seeing – and realizing – what previously was largely unheeded. There are not two structures but many co-existing structures with diverse shapes. The number of molecules that populate a distinct shape (with a distinct set of coordinates) is determined by their relative energies. Protein folding diagrams have been portraying the different states down the funnel, aiming at the global minimum, under the assumption that this is the state that executes the function. Less attention was paid to the many states around the native basin. The Frauenfelder, Sligar, and Wolynes portrayal helped inspire a function-oriented perspective (Nussinov and Tsai, Reference Nussinov and Tsai2014a). The protein folding enigma is still unresolved (Dill et al., Reference Dill2007; Montgomery Pettitt, Reference Montgomery Pettitt2013; Chen et al., Reference Chen2023; Straiton, Reference Straiton2023); it has been dealt with by sidestepping it, through the enormously useful predictive software and databases (Baek et al., Reference Baek2021; Jumper et al., Reference Jumper2021; Ohno et al., Reference Ohno2024; Varadi et al., Reference Varadi2024). The free energy landscape of proteins (and broadly biomacromolecules) is powerful since it captures on a two (or three)-dimensional Cartesian coordinate system all possible shapes that protein molecules can populate as a function of their corresponding energy levels. This landscape idea is compelling: it conveys that in solution, the protein may populate any of these shapes at the same time, albeit with different likelihoods. On the downside, not conveyed by this static portrayal is that function is not about reaching a ‘passive’ equilibrium in solution, but about homeostasis in a living cell. For the population to respond to a change in the cellular conditions, including incoming signals, the relative propensities of the shapes must be dynamically – actively – regulated, commonly by an allosteric effector. Effectors include proteins, lipids, water molecules binding at specific sites, ions, ligand concentration, and temperature. Function is about biasing the ensembles by effectors since the stability of the protein molecules is altered by the binding events, and broadly, the physico-chemical environment (Boehr et al., Reference Boehr, Nussinov and Wright2009; Aykut et al., Reference Aykut2013; Wei et al., Reference Wei2016; Biddle et al., Reference Biddle2021; Janson et al., Reference Janson2023; Nussinov et al., Reference Nussinov2023b), that is, function is executed by dynamic ensembles whose redistribution is triggered by allosteric events. Such redistributions take place under physiological conditions and in disease by mutations, overexpression, and drugs. This crucial function-related rendering has not been intimated by the original landscape plots. In the decades since, exactly how these redistributions take place has been the center of many studies, theoretical, computational, and experimental (Di Pietrantonio et al., Reference Di Pietrantonio2019; Janson et al., Reference Janson2023; Yabukarski, Reference Yabukarski2024; Flores et al., Reference Flores2025).

We were inspired by this Frauenfelder, Sligar, and Wolynes perspective, but missed the biological, functional connection, leading us to suggest in the 1990s a dynamic ensemble landscape concept (Ma et al., Reference Ma1999; Tsai et al., Reference Tsai1999a, Reference Tsai1999b; Kumar et al., Reference Kumar2000; Gunasekaran et al., Reference Gunasekaran2004; Nussinov et al., Reference Nussinov2019a). We proposed that since different conformations can better execute distinct functions, the landscape needs to epitomize the dynamic response through the shifts of the population. Among the many different shapes and surface chemistries, there are expected to be certain ones that fit the incoming ligand better. We then suggested that these are ‘selected’, triggering the redistribution of the ensemble, propagation of the allosteric signal, and cell function (Figure 1). Since the relative stability, thus population, of that shape may be low, allosteric propagation overcoming the kinetic barriers with optimization may be slow. If, however, the ligand concentration is very high, induced fit may prevail over conformational (or shape) selection (Weikl and von Deuster, Reference Weikl and von Deuster2009; Kar et al., Reference Kar2010). Because the specific protein shape often has a low population, which will require crossing barriers, induced fit has faster timescales, thus more readily detected experimentally and computationally by simulations (Boehr et al., Reference Boehr, Nussinov and Wright2009), clarifying the often-mistaken conclusion of an induced fit mechanism (Kumar et al., Reference Kumar2000; McClendon et al., Reference McClendon2009). Nuclear magnetic resonance (NMR) data detail how transitions between high-energy states can take place (Gardino et al., Reference Gardino2009). Especially, induced fit is likely to encounter kinetic blocks, leading Hans Bosshard to wonder “molecular recognition by induced fit: how fit is the concept?” (Bosshard, Reference Bosshard2001). Induced fit mechanisms can be more clearly thought of in the context of standard free energy surface diagrams used in Figure 1, which can incorporate both kinetic and equilibrium aspects of the protein networks involved in terms of the barrier heights between different interconverting protein states.

Allostery and regulation are rooted in population shift, which is how constitutive allosteric driver mutations, allosteric post-translational modifications (PTMs), signaling lipids, and ions can work (Nussinov and Tsai, Reference Nussinov and Tsai2013; Nussinov et al., Reference Nussinov2013b; Tsai and Nussinov, Reference Tsai and Nussinov2013). Changes in conditions, such as temperature, salinity, pH, and membrane environment, work by altering the stabilities of specific chemical interactions, thus their competitiveness and the relative stabilities of the conformations. An increase in the ligand (e.g., drug) concentration works by increasing the probability of a conformational change, thereby influencing function. In all scenarios, allosteric events shift the populations rather than creating new ones (Tsai et al., Reference Tsai1999a, Reference Tsai1999b; Boehr et al., Reference Boehr, Nussinov and Wright2009; del Sol et al., Reference del Sol2009). These shifts are allosteric bias.

Allosteric signals propagate through interactions, coupling allosteric and functional sites (Sadovsky and Yifrach, Reference Sadovsky and Yifrach2007; Buchenberg et al., Reference Buchenberg2017; Naganathan, Reference Naganathan2019; Jang et al., Reference Jang2020; Wen et al., Reference Wen2022; Diaz et al., Reference Diaz2023; Poudel and Leitner, Reference Poudel and Leitner2023). “How allostery works” (Tsai and Nussinov, Reference Tsai and Nussinov2014) reasons that allosteric communication is established by networks through which strain energy generated at the allosteric site by an allosteric event propagates to the functional site, promoting a conformational and dynamic change (Buchenberg et al., Reference Buchenberg2017; La Sala et al., Reference La Sala2017; Verkhivker et al., Reference Verkhivker2020; Chen et al., Reference Chen2022; Haliloglu et al., Reference Haliloglu2022; Mugnai and Thirumalai, Reference Mugnai and Thirumalai2023; Samanta et al., Reference Samanta2023; Marfoglia et al., Reference Marfoglia2024). While sometimes misconstrued, in general, the more distant the allosteric and functional sites in the structure, the less productive the allosteric effect. This is because the propagating signal needs to cross more allosteric transitions. More competent allosteric sites are commonly closer.

Dynamic communication networks and functional specificity

Allosteric networks: Examples

To provide examples, in one sagacious work, allosteric effects have been delineated by machine learning methods, including deep neural networks and random forests, applied to MD simulations, and represented as residue-specific properties in “residue response maps” (Hayatshahi et al., Reference Hayatshahi2019). As Liu and her colleagues note, these vindicate allostery as a residue-specific concept. By quantitatively varying attributes, “all residues could be considered as allosteric residues because each residue ‘senses’ the allosteric events.” These maps could serve as fingerprints of allosteric consequences upon residue perturbation. In another example, an innovative network-based perturbation strategy discovered connections between allosteric centers and binding hotspots of epistatic couplings (Raisinghani et al., Reference Raisinghani2024). In a PDZ domain example (Figure 2a), altered binding affinities associated with allosteric communication through different pathways were identified by perturbation response scanning (Gerek and Ozkan, Reference Gerek and Ozkan2011). Changes of the allosteric network were shown by the dynamic coupling of different residue pairs.

Figure 2.

Altered allosteric networks and pathways mapped onto protein structures. (a) Allosteric networks highlighted in the crystal structure of the PDZ3 domain of the 95 kDa postsynaptic density protein, PSD-95 (PDB ID: 1BFE). In PDZ3 with a truncated C-terminal region, the allosteric pathway differs. Data for PDZ3 allosteric networks were obtained from the literature (Gerek and Ozkan, Reference Gerek and Ozkan2011). (b) Solution NMR structure of the Abl kinase domain (PDB ID: 6XR6) in the active state. In allosteric communication, the dynamically coupled residues in the hinge cluster, the DFG motif, and the R-spine are marked on the structure (Krishnan et al., Reference Krishnan2022). The transparent surface in the C-lobe kinase domain indicates the allosteric binding pocket for the myristoyl group or allosteric inhibitor. (c) Crystal structures of Src kinase in the inactive state (PDB ID: 2SRC) and in the active state (PDB ID: 1Y57). The inactive Src shows a closed conformation within autoinhibition, while the active Src shows an open conformation. The dynamically coupled residues in the active site are marked on the structures (Foda et al., Reference Foda2015). (d) Crystal structures of the GDP-bound K-Ras4B with the G12D (PDB ID: 5US4), K104Q (PDB ID: 6WS2), and G12D/K104Q (PDB ID: 6WS4) mutations. The mutations allosterically regulate nucleotide exchange by the guanine nucleotide exchange factor (GEF). The residues involved in the allosteric network are marked in the structures (Yang et al., Reference Yang2023). The mutant residues are highlighted in red. Abbreviations: SI, Switch I; SII, Switch II.

MD and ensemble-based distance fluctuations analysis of Abl kinase revealed dynamic allosteric networks between the ATP site, the substrate binding region, and the allosteric binding pocket in the active and inactive states (Krishnan et al., Reference Krishnan2022). The networks differ, with perturbation-based network analysis capturing their shifts. The dynamic coupling between residue N387, the DFG motif, and the R-spine mediates structural stability and allosteric communications (Figure 2b). Especially, the long-range communication paths between the ATP binding site and the substrate regions in the active state were suppressed in the closed inactive state and replaced with the stronger allosteric couplings between the ATP site and the allosteric binding pocket. The allosteric coupling between the sites in the inactive state and the network shifts are consistent with NMR data of Kalodimos and his colleagues, who determined the stable active state and two short-lived ligand-free conformations occurring only in 5% of the time (Foda et al., Reference Foda2015; Xie et al., Reference Xie2020; Krishnan et al., Reference Krishnan2022).

Seeliger and his colleagues provided remarkable insight into switching Src kinase networks (Foda et al., Reference Foda2015). Src is a non-receptor cytoplasmic protein tyrosine kinase (PTK) (Figure 2c). The highly concerted conformational change observed in their MD simulations suggested dynamically coupled networks, with ATP and peptide substrates exhibiting negative – and ADP and phosphorylated peptides positive cooperativity. Protonation of the DFG motif plays a central role in regulating the conformational switching (Bukhtiyarova et al., Reference Bukhtiyarova2007; Jacobsen et al., Reference Jacobsen2012; Lovera et al., Reference Lovera2012). Mg²⁺ and ATP (and ADP) favor active DFG conformation. Following phosphoryl transfer, the protonation state and DFG conformation are transposed.

Allosteric regulation of Switch II domain controls K-Ras oncogenicity (Yang et al., Reference Yang2023). G12D and K104Q mutations arrest nucleotide exchange by the guanine nucleotide exchange factor (GEF). K104 interactions were observed to be essential for K-Ras oncogenicity, being part of an allosteric network with M72, R73, and G75 on the α2-helix of the Switch II region (Figure 2d). Its mutation switches networks.

Finally, binding of specific ligands to GPCRs drives distinct signal transduction and conformational transitions. Binding of the C–C motif chemokine ligands 19 (CCL19) and 21 (CCL21) to the C–C chemokine receptor type 7 (CCR7) triggers molecular switches through altered side-chain conformations, leading to concerted transmembrane helical domain motions and distinct receptor conformational states (Gaieb et al., Reference Gaieb2016; Gaieb and Morikis, Reference Gaieb and Morikis2017) (Figure 3). The conformationally inhomogeneous ligand-free apo-state samples these states. This example links network switching to signaling bias. Importantly, as above, while we discuss switching between two conformations, multiple functionally relevant conformations are visited. A two-conformation scenario cannot explain the positive binding cooperativity of phosphopeptide with both ADP and ATP (Foda et al., Reference Foda2015), nor the other scenarios discussed above.

Figure 3.

Biased GPCR signaling. Binding of different chemokine (C-C motif) ligands, CCL19 and CCL21, to CCR7 results in biased signaling (top panel). CCL19 activates both the G-protein and β-arrestin signaling pathways. While CCL21 activates G-protein signaling, it activates β-arrestin signaling less robustly. CCL19/CCR7 activates both G protein-coupled receptor kinases, GRK3 and GRK6, while CCL21/CCR7 activates GRK6. GRKs phosphorylate the C-terminal region of CCR7, resulting in the recruitment of β-arrestin, which ultimately leads to receptor desensitization and internalization. G-protein signaling regulates the PI3K/AKT, MAPK (ERK, p38, JNK), and RhoA/cofilin pathways, leading to cell survival, chemotaxis, and migration, respectively. Crystal structures of CCL19 (PDB ID: 7STA), CCL21 (PDB ID: 5EKI), and CCR7 (PDB ID: 5EKI) with arrows indicating ligand binding on the extracellular side. Schematic representation of biased signaling by agonist binding to GPCRs (bottom panel). Agonist ligand binding to GPCRs triggers a conformational change in the receptor, leading to receptor activation and subsequent interaction and activation of heterotrimeric G proteins. During receptor activation, GRKs phosphorylate the C-terminal region of GPCRs, promoting β-arrestin binding. With a balanced (unbiased) agonist, GPCRs can activate both G-protein and β-arrestin signaling. However, with biased agonists that bind at different positions on the receptor, GPCRs activate either G-protein or β-arrestin signaling depending on the type of agonist ligand. Biased receptor signaling occurs for GPCRs that lack C-terminal phosphorylation sites for β-arrestin recruitment. In this case, the receptors activate only G-protein signaling. Different levels of expression for signaling effectors result in biased system signaling. In this example, β-arrestin signaling is enhanced by highly expressed GRKs and β-arrestins. Schematic diagrams adapted concept from literature (Smith et al., Reference Smith2018).

Signaling bias not only in GPCRs but also in all dynamic, allosteric proteins

Through conformational biasing, allostery governs signaling, and signaling dominates cell behavior (Nussinov et al., Reference Nussinov2012; Nussinov et al., Reference Nussinov2013a). Signaling bias is an intrinsic property of all dynamic proteins, since all harbor conformational ensembles (Nussinov et al., Reference Nussinov2014). Thus, rather than all-or-none conformational states, this fosters a view of allostery in terms of the statistics of the multiple states, with changes in the statistics biased by allosteric agents. Because each state can have a distinct function, altered statistics result in biased functional change. The changes in the statistics of the states reflect changes in their relative stabilities and dynamics, which are a function of the interactions that they maintain. The interactions can be expressed by small or large conformational (enthalpic) and dynamic (entropic) changes (Tsai et al., Reference Tsai2008). Sampling conformational space along allosteric propagation pathways is not homogeneous, making biasing a probabilistic event. Lower barriers between populated states are favored, resulting in faster time scales (Ma et al., Reference Ma2011). The ensembles of the native protein populate multiple local basins. Those of the agonist-biased states are among them. Pre-agonists’ binding, these basins are sparsely populated. Agonist-mediated signaling to the functional sites biases the ensembles, increasing the respective basins’ populations (Tsai and Nussinov, Reference Tsai and Nussinov2014). Agonists redistribute the ensembles by favoring pathways that are conducive to their signaling. Different agonists bind at different allosteric sites. The favored pathway that each activates depends on the specific frustration event triggered by the specific allosteric agonist at the specific site (Nussinov and Tsai, Reference Nussinov and Tsai2014b).

While the literature is enriched with single allosteric events in single proteins, in reality, multiple simultaneous allosteric events take place on the same protein (or complex (Abbate et al., Reference Abbate2024) and impact not only the protein but signaling in living cells (Mayer, Reference Mayer1999; Gunasekaran et al., Reference Gunasekaran2004; del Sol et al. Reference del Sol2009; Kar et al. Reference Kar2010; Korcsmaros et al., Reference Korcsmaros2010; Nussinov, Reference Nussinov2013; Nussinov et al., Reference Nussinov2013a; Suderman and Deeds, Reference Suderman and Deeds2013; Smith et al., Reference Smith2016; Chiesa et al., Reference Chiesa2020; Nussinov et al., Reference Nussinov2021; Prischi and Filippakopoulos, Reference Prischi and Filippakopoulos2021; Xia et al., Reference Xia2021; Morris et al., Reference Morris2022; Klumpe et al., Reference Klumpe2023; Liao et al., Reference Liao2023; Greenblatt et al., Reference Greenblatt2024). Protein assemblies (Hamley, Reference Hamley2019; Chiesa et al., Reference Chiesa2020; Olechnovic et al., Reference Olechnovic2023; Pillai et al., Reference Pillai2024; Shor and Schneidman-Duhovny, Reference Shor and Schneidman-Duhovny2024; Wang et al., Reference Wang2025), and specifically scaffolding proteins (Smith and Scott, Reference Smith and Scott2013; Perry et al., Reference Perry2019; DiRusso et al., Reference DiRusso2022; Yerramilli et al., Reference Yerramilli2022; Kim et al., Reference Kim2023; Thines et al., Reference Thines2023), provide ultimate examples of allosteric multi-protein associations. The observed cell responses are the ramification of all. Cell activities, signaling, and catalysis (Tsunoda et al., Reference Tsunoda1998; Wang et al., Reference Wang2014; Sherman et al., Reference Sherman2016; Ha et al., Reference Ha2020; McKee et al., Reference McKee2025), are biased by integration of all events, although exactly how has not been explored, despite such occurrences being ubiquitous. Signaling proteins commonly interact with multiple partners and their actions are shaped by multiple allosteric effectors (Tsai et al., Reference Tsai2009). Such integrated shape-biasing scenarios can explain temporally favored partner selection by a multi-partner protein (Nussinov et al., Reference Nussinov2013a). Effective local concentration of a partner plays a role by increasing the probability of its binding.

Conformational switches have been observed, for example, in the allosteric regulation of autoinhibition and activation of c-Abl (Liu et al., Reference Liu2022), a non-receptor tyrosine kinase, by the myristoyl group, a fatty acid PTM, which biases Abl’s ensembles toward autoinhibition and activation by serving as a mechanical lever, is one example. c-Abl regulates cell growth and survival in healthy cells. It also promotes chronic myeloid leukemia when its N-terminal region is linked to the C-terminal region of the Bcr. The myristoyl group binds to a hydrophobic pocket in the kinase domain’s C-lobe. Its interactions bias the population toward a kink in the kinase C-terminal αI-helix (de Buhr and Grater, Reference de Buhr and Grater2023). The resulting docking of the SH2 and SH3 domains to the kinase C lobe reduces its conformational flexibility, inducing changes in the active site, triggering the autoinhibited state. Myristoyl’s dissociation from the kinase domain pocket triggers SH2-SH3 release and activation.

Another example involving c-Myc’s population bias also relates to stabilized (destabilized) states (Boi et al., Reference Boi2023). c-Myc’s transcript is highly unstable (half-life of 20–30 min) (Dani et al., Reference Dani1984). The protein’s turnover timescale is 25 min (Hann and Eisenman, Reference Hann and Eisenman1984), with c-Myc’s acetylation and methylation playing a role (Farrell and Sears, Reference Farrell and Sears2014), and phosphorylation and dephosphorylation at S62 and T58 for c-Myc (isoform p64 or c-Myc2) being the prime function and degradation orchestrators (Figure 4). S62-phosphorylated c-Myc associates with MAX (Myc-associated factor X). With other E-box components, it drives transcription of target genes. The ensuing phosphorylation event at T58 results in the removal of the phosphate from S62, destabilizing c-Myc, leading to its tagging with ubiquitin and degradation (Boi et al., Reference Boi2023). Broadly, binding events may weaken or favor other interactions, triggering homeostasis pathway response. If allostery is not at play ensemble ensemble-biased signal propagation will not take place.

Figure 4.

Relative stability of c-Myc by phosphorylation. Ras activates the MAPK and PI3K/AKT/mTOR pathways, which upregulate c-Myc (top panel). ERK1/2 and GSK3β phosphorylate Ser62 and Thr58 in c-Myc, respectively, leading to increased stability of c-Myc. The phosphatase PP2A activates GSK3β by dephosphorylation, whereas AKT deactivates it by phosphorylation. PP2A dephosphorylates pSer62 in c-Myc, leading to destabilization of c-Myc only with pThr58 and resulting in proteasomal degradation. Molecular structures of c-Myc (blue cartoon) and MAX (red cartoon) transcription factors (TFs) predicted by AlphaFold (bottom panel). Crystal structure of their assembled complex recognizing DNA (PDB ID: 1NKP). The phosphorylation sites S62 and T58 are highlighted. For c-Myc, the sequence refers to c-Myc2, the isoform p64 (UniProt ID: P01106).

GPCRs and dynamic allosteric proteins share signaling principles

The GPCR literature is rich with multiple excellent reviews [e.g., (Sriram and Insel, Reference Sriram and Insel2018; Gurevich and Gurevich, Reference Gurevich and Gurevich2020; Sutkeviciute and Vilardaga, Reference Sutkeviciute and Vilardaga2020; Kenakin, Reference Kenakin2021; Lu et al., Reference Lu2021; Yang et al., Reference Yang2021; Hedderich et al., Reference Hedderich2022; Huang et al., Reference Huang2023; Schmidt, Reference Schmidt2023; Zhang et al., Reference Zhang2024; Borah et al., Reference Borah2025; Conflitti et al., Reference Conflitti2025; Ma et al., Reference Ma2025; Pearce et al., Reference Pearce2025)]. The reports illustrate that the principles of the signaling bias observed in GPCRs are not GPCRs-specific. They are shared by all dynamic allosteric proteins (Isogai et al., Reference Isogai2016; Solt et al., Reference Solt2017; Yuan et al., Reference Yuan2018; Raniolo and Limongelli, Reference Raniolo and Limongelli2020; Huang et al., Reference Huang2021; Kenakin, Reference Kenakin2021; Di Marino et al., Reference Di Marino2023; Huang et al., Reference Huang2023; Vogele et al., Reference Vogele2023; D’Amore et al., Reference D’Amore2024; Jones et al., Reference Jones2024). It is the functional and pharmacological significance of GPCRs, that led to the wide association of signaling bias with GPCRs and increasingly also with RTKs [e.g., (Karl et al., Reference Karl2020; Trenker and Jura, Reference Trenker and Jura2020; Wirth et al., Reference Wirth2023; Gross et al., Reference Gross2024; Ma et al., Reference Ma2025)]. In line with our discussion above, biophysical studies observed that ligands with distinct bias profiles stabilize distinct GPCR conformations (Liu et al., Reference Liu2012; Wingler et al., Reference Wingler2019; Suomivuori et al., Reference Suomivuori2020). Signaling bias is a fundamental, inherent property. This is also held for other GPCR mechanics. As put forward by Smith et al. (Smith et al., Reference Smith2018), in GPCRs, biased signaling may be determined by the receptor (receptor bias) and by the relative expression levels of transducers (system bias) (Figure 3). Again, here too, the observation that different ligands, or allosteric effectors, activate (or suppress) different cellular pathways, is not unique to GPCRs. Further, it has recently been shown that shifts in populations of minor conformational states can impact GPCRs’ function (Suomivuori et al., Reference Suomivuori2020; Vogele et al., Reference Vogele2025). These are well-established scenarios in other proteins. The binding of different extracellular ligands of angiotensin II type 1 receptor (AT1R) results in distinct GPCR conformations that favor different signaling partners (Gq protein α subunit and β-arrestins) (Wingler et al., Reference Wingler2019). These shifts take place because GPCRs, like all dynamic proteins, embody ensembles. As to RTKs (Freed et al., Reference Freed2017; Karl et al., Reference Karl2020; Kiyatkin et al., Reference Kiyatkin2020), the fibroblast growth factor and its receptor system (FGF/FGFR) is of special interest (Krzyscik et al., Reference Krzyscik2024). With 18 FGF ligands and four FGFRs, even with a single FGFR expressed, different FGFs can trigger dramatically different quantitative and qualitative responses. Further, expression levels can influence biased signaling not only in GPCRs (Smith et al., Reference Smith2018), nuclear hormone receptors, and RTKs, but commonly proteins involved with malignancy. Overexpression of oncogenic proteins is often observed in aggressive cancers (Nussinov et al., Reference Nussinov2025). Early on, they were dubbed ‘brake-override’ proteins (Liberal et al., Reference Liberal2012) that enable the development of some cancers. The higher expression level results in more protein molecules, thus more active conformations. In cancer, these break normal physiological bias, shattering signaling guardrails, leading to dedifferentiation (Nussinov, Reference Nussinov2025).