Introduction: The old paradigm and its transformation

Nussinov and Wolynes questioned whether the energy landscapes of biomolecular function constituted a second molecular biology revolution (Nussinov and Wolynes, Reference Nussinov and Wolynes2014). The old sequence–structure–function paradigm of molecular biology posited that a protein’s unique three-dimensional structure is vital for its biological function. This idea, which assumed the existence of a single stable shape, was the cornerstone of molecular biology for decades. However, emerging physics-based inconsistencies have challenged this classical model, suggesting that the concept of a single, rigid shape may be an untenable and ill-conceived premise (Nussinov and Wolynes, Reference Nussinov and Wolynes2014). To what extent did the ascendance of the sequence-dynamic structure–function paradigm represent a second, Kuhnian scientific revolution in molecular biology (Nussinov et al., Reference Nussinov2023)? Kuhnian revolutions are triggered by accumulated anomalies that the old paradigm cannot explain (Kuhn, Reference Kuhn2012), which has been the case in molecular biology. Molecular biology struggled to explain how static protein molecules work (Raman et al., Reference Raman1992; Foote and Milstein, Reference Foote and Milstein1994; Zeder-Lutz et al., Reference Zeder-Lutz1994; Berger et al., Reference Berger1999; Orgel et al., Reference Orgel2000; Bosshard, Reference Bosshard2001; Nussinov, Reference Nussinov2025). It also struggled to explain how a rigid protein molecule can bind to a drug when its active site structure, seen in the crystal, does not fit the drug’s shape, which prompted the ‘induced fit’ doctrine (Koshland, Reference Koshland1958; Finer-Moore et al., Reference Finer-Moore1997; Yu and Koshland, Reference Yu and Koshland2001). Molecular biology followed a dogma of ‘what is not observed does not exist’, which still lingers today (Boehr et al., Reference Boehr2009). If only one structure (or perhaps two: ligand-bound and unbound) is observed in crystal structures, it was assumed only that single form (or those two forms) existed. Yet, as Frauenfelder et al.’s (Reference Frauenfelder1991) powerful, transformational physics-based idea indicated, a protein molecule can have many shapes, whose varied energies determine their populations (Bryngelson and Wolynes, Reference Bryngelson and Wolynes1987; Bryngelson et al., Reference Bryngelson1995; Onuchic et al., Reference Onuchic1995; Onuchic et al., Reference Onuchic1996; Panchenko et al., Reference Panchenko1996; Onuchic et al., Reference Onuchic1997; Shoemaker et al., Reference Shoemaker1997; Hardin et al., Reference Hardin2002). Their breakthrough idea was a significant departure from earlier, simpler models, reshaping the understanding of protein dynamics. Still, it took years for the community to overcome its initial resistance to fully recognize its validity and significance, finally becoming a widely accepted paradigm (Tsai et al., Reference Tsai1999b; Nussinov, Reference Nussinov2025). Its landmark implications were demonstrated to be immense, much beyond the protein folding problem, which was its initial aspiration. As imagined by Nussinov and her colleagues, its reverberations embrace practically all dynamic life processes at the molecular level, including molecular recognition, catalysis, and, broadly, cell regulation (Nussinov et al., Reference Nussinov2023).

Thus, while some in the community initially resisted change and dismissed physics-based inconsistencies, eventually, the energy landscape theory emerged, replacing the old and triggering new understanding and progress (Schon, Reference Schon2024). Rather than the information flow of the sequence–structure–function paradigm, the molecular biology revolution that prevailed heralded the sequence-dynamic structure–function paradigm (Nussinov et al., Reference Nussinov2023), reshaping biology, and merited the designation of a second revolution (Nussinov and Wolynes, Reference Nussinov and Wolynes2014; Nussinov et al., Reference Nussinov2019). The Conformational Selection and Population Shift paradigm (Ma et al., Reference Ma1999; Tsai et al., Reference Tsai1999a; Boehr et al., Reference Boehr2009; Nussinov, Reference Nussinov2025) demonstrated how the energy landscapes metamorphose into dynamic regulated biological signals, thereby revolutionizing the understanding of binding, allostery, and signaling.

Below, we briefly overview the free energy landscape theory within the framework of molecular biology, first focusing on its protein folding origin and continuing with free energy landscapes and their functional message, which is our focus. Following these introductory sections, we pivot to our crucial core thesis that dynamic macromolecules execute their function by conforming to the laws of physics. We discuss the landscapes of disordered protein states, whose electrostatic interactions, along with a lack of a strong hydrophobic core, typically result in multiple shallow energy minima separated by low kinetic barriers, enabling rapid interconversion between many different conformations. Focusing on barriers and transition states, we suggest that proteins hopping from one conformation to the other is a key to their function. Finally, for a comprehensive outlook on landscapes of biomacromolecules from the conformational and functional standpoints, we discuss the complex free energy landscapes of RNA. Collectively, free energy landscapes are fundamental to understanding protein function and cell life, challenging the obsolete paradigm of single, rigid protein shapes, contributing to scientific transformation.

Conformational selection and population shift: The paradigm

Unlike the ‘induced fit’ model, where a ligand forces a protein into a new shape, the ‘conformational selection and population shift’ paradigm posits that all functional conformations already pre-exist in a dynamic equilibrium (Figure 1). A ligand simply ‘selects’ and stabilizes a complementary state from this pre-existing pool. When a ligand binds and stabilizes a specific conformation, the entire ensemble re-equilibrates. This re-equilibration ‘shift’ increases the relative population of the binding-favored state, which is the physical origin of the allosteric effect.

Figure 1.

Conformational selection illustrated by free energy landscapes of protein folding and a conformational ensemble. (A) A funnel-shaped free energy landscape of protein folding. The lowest free energy region at the bottom contains multiple native-like functional conformational states. (B) A magnified view of the funnel bottom showing a conformational ensemble containing multiple substates with small structural differences, separated by low kinetic barriers (left). A ligand selectively binds to a pre-existing complementary conformation from the native ensemble, stabilizing that specific conformation, making it more energetically favorable and shifting the population distribution toward the ligand-bound substate (right).

To understand dynamic propagation and how it relates to ‘induced fit’, we consider an action movie. Under standard projection protocol, we see movement. However, if the projection is slowed, we see a series of still-like states. For proteins, this series of still-like states is a series of conformations. Rolling the movie tape faster, the conformational transitions may not be seen, but their subsequent outcome is. Coming back to Koshland’s ‘induced fit’ (Koshland, Reference Koshland1994), we can envision it as not seeing the distinct protein hopping over barriers between states, thus seeming to have a ligand ‘pushing’ or ‘pulling’ the protein to reach the optimal state. Such a depiction can bridge barriers between the old and the new paradigms. Regardless, Koshland’s model initiates from the Key–Lock theory, whereas the conformational selection and population shift initiates from the most compatible ‘selected’ state, which might be sparsely populated. Thus, in the energy landscape, its basin may not be deep. However, ligand binding will reshape landscape basins and barriers. This is how the conformational ensemble (the collection of all possible protein shapes) is dynamically modulated by the ligand, which leads to the observable ‘induced fit’ effect. Under an ensemble regime, the outcome is a shift in the population, not the creation of a brand-new shape from scratch.

The energy landscape theory and protein folding

Peter Wolynes and his collaborators (José Onuchic, Joseph Bryngelson, Zaida Luthey-Schulten, and others) developed the statistical mechanical energy landscape theory, which provided a rigorous theoretical framework for understanding the rugged, funnel-shaped nature of the energy surface (Figure 1). They proposed that the free energy landscape of a foldable protein is not a random rugged surface, but a ‘funnel’ (Bryngelson and Wolynes, Reference Bryngelson and Wolynes1987; Frauenfelder et al., Reference Frauenfelder1991; Bryngelson et al., Reference Bryngelson1995; Wolynes et al., Reference Wolynes1995; Onuchic et al., Reference Onuchic1997; Wolynes, Reference Wolynes2015). They showed that different than poorly behaving glass transitions (Frauenfelder et al., Reference Frauenfelder1991), folding funnels guide the protein toward its single, most stable, functional native state through many possible but biased pathways, thus explaining how folding can occur on a biological timescale, resolving Levinthal’s paradox. They proposed the principle of minimal frustration, which suggests that evolved proteins are optimized by natural selection to minimize conflicts (frustration) in achieving their native structure, making the folding process thermodynamically and kinetically favorable. This principle is crucial because it posits that naturally evolved protein sequences are specifically optimized to minimize conflicts during folding, ensuring that correct structures are stabilized (Schon, Reference Schon2024).

This landmark energy landscape work (Frauenfelder et al., Reference Frauenfelder1991) followed Frauenfelder’s discovery of conformational substates in his studies of myoglobin carbon monoxide (Austin et al., Reference Austin1975). As noted by Bill Eaton, this landscape work ‘played a major role in demonstrating the power and importance of dynamics and of Kramers reaction rate theory for understanding protein function. The biggest impact of energy landscape theory has been in the protein folding field, which is well-known and has been documented in numerous articles and reviews’ (Eaton, Reference Eaton2021). The energy landscape theory has been considered a foundational concept in the modern understanding of protein folding kinetics and thermodynamics. In this review, we add its monumental influence on function.

While sharing the funnel concept, Dill emphasized the hydrophobic effect and the role of conformational entropy in constraining the vast number of possible protein shapes, largely focusing on how the compactness of a polymer in a solution drastically reduces the search space for the native structure, helping resolve Levinthal’s paradox (Dill Reference Dill1985; Reference Dill1990; Dill et al., Reference Dill1995; Dill and Chan, Reference Dill and Chan1997).

The free energy landscape and protein function

In the 1990s, the biological thinking of Nussinov and her colleagues focused on the functional message of the free energy landscape postulated by Frauenfelder, Sligar, and Wolynes (Frauenfelder et al., Reference Frauenfelder1991), pioneering the concept of leveraging energy landscapes to elucidate functional mechanisms on the structural level (Nussinov, Reference Nussinov2025), demonstrating that the dynamic nature is fundamental to molecular function. The energy landscape was conceptualized as a dynamic continuum, where conformational shifts drive biological signaling and function, showcasing their immense explanatory power and potential. Landscape basins depict conformational stability, serving as candidate drug targets; hills separating them represent their transitions, collectively capturing biological processes like allosteric communication (Ma et al., Reference Ma1999; Tsai et al., Reference Tsai1999a; Tsai et al., Reference Tsai1999b; Kumar et al., Reference Kumar2000; Gunasekaran et al., Reference Gunasekaran2004), activation, and autoinhibition. Minimizing frustration is also key in allosteric communication, and broadly in the regulation of biological processes (Ferreiro et al., Reference Ferreiro2014; Gianni et al., Reference Gianni2021; Nussinov et al., Reference Nussinov2023; Nussinov, Reference Nussinov2025; Parra et al., Reference Parra2025). However, whereas fast protein folding rates can minimize errors, fast transitions among conformations as the allosteric signal propagates are not necessarily a functional aim. Instead, the rate may be associated with function. Processes can be slowed down by reduced protein translation rates, resulting in exiting the cell cycle in differentiation (Nussinov et al., Reference Nussinov2024). They can also be influenced by the rate of protein state conversion reminiscent of ribosome stalling by synonymous single-nucleotide polymorphisms (SNPs) in multidrug resistance 1 protein (MDR1 or P-glycoprotein 1 encoded by ABCB1) (Figure 2). A synonymous (rare) codon substitution was proposed to be able to lead to altered folding pathways and distinct minima by fold-switching through kinetically trapped states (Tsai et al., Reference Tsai2008b). There, frequent-to-rare codon substitutions have led to identical protein sequences. However, the long-enough ribosomal pause timescales may have resulted in altered structure and function of MDR1. For some functions, or under certain conditions, slower could be better (Wayment-Steele et al., Reference Wayment-Steele2024).

Figure 2.

Synonymous single-nucleotide polymorphisms (SNPs) in ABCB1. The ABCB1 gene encodes MDR1 (also known as P-glycoprotein 1), a critical membrane pump that removes foreign substances from cells. Synonymous SNPs in ABCB1 are often labeled ‘silent’ because they do not alter the amino acid sequence of P-glycoprotein 1 (top panel). However, the 1236C > T and 3435C > T polymorphisms in ABCB1 demonstrate that these changes are not always functionally neutral. By altering cotranslational folding and membrane insertion timing, these SNPs create alternative protein conformations with distinct substrate-binding affinities. This proves that genetic variations can fundamentally change drug interactions without changing a single amino acid. Membrane-bound P-glycoprotein 1 is a symmetrical transporter composed of two homologous halves (bottom panel). Each half contains an extracellular domain (ECD), a transmembrane domain (TMD), which forms the substrate translocation pathway, intracellular domain (ICD), and a cytoplasmic nucleotide-binding domain (NBD) that provides the ATP-binding pocket. Polymorphisms affect the rate of cotranslational folding and membrane insertion, subtly altering the final structure of these binding pockets. In the apo state (ligand-free), P-glycoprotein 1 adopts a closed inward-facing conformation that covers the substrate-binding site (PDB ID: 9CR8). Binding of a substrate, such as Taxol, to the TMD induces a conformational shift, yielding an open inward-facing conformation that increases NBD affinity for ATP (PDB ID: 9CTF). Upon ATP binding, the protein transitions to an open outward-facing conformation. This reduces substrate affinity and releases the molecule into the extracellular space. Substrate release subsequently leads to a closed outward-facing conformation (PDB ID: 9CTG). Finally, ATP hydrolysis resets the transport cycle. This P-glycoprotein 1 conformational cycle mechanism in the membrane was adopted from the literature (Kurre et al., Reference Kurre2025). Part of graphics created with BioRender.

Dynamic molecules conform to the laws of physics

Transformative periods in molecular biology have been driven by physical and chemical sciences. One revolutionary idea, based on physical evidence and an understanding of chemical structure, was the discovery of the DNA double helix in 1953. A second physics- and chemistry-based revolution advanced the avant-garde idea of the free energy landscapes of biomolecular function (Nussinov and Wolynes, Reference Nussinov and Wolynes2014; Nussinov et al., Reference Nussinov2019). This powerful structure–function idea convincingly argues that to function, dynamic macromolecules must always interconvert between a variety of structures with varying energies. Static molecules cannot execute (dynamic) functions; only dynamic molecules that conform to the laws of physics and chemistry can. This idea compellingly explained that the common ‘dynamic molecules’ description does not simply mean molecules undergoing ‘jiggling and wiggling’, as famously used by Richard Feynman (Tantillo, Reference Tantillo2018). Instead, the breakthrough idea was that these dynamics involve proteins transitioning between pre-existing conformations, where the transitions depend on the relative energies of the conformations (minima basins in the landscape) and the kinetic barriers (the hills) that separate them, as portrayed by free energy landscapes (Figure 1). The free energy landscape explains that protein dynamics are caused by transitions, or ‘leaping’ between conformations, and that this ‘hopping’ dictates protein function. Within each of these conformational states, the protein is not static. It may undergo continuous, localized fluctuations between substates (microstates), facilitating its transitions. How fast the protein leaps between states is a key factor in how fast the allosteric communication pathway flows, and how fast the biological process proceeds. This foundational understanding provides a unified approach to relate protein sequence, structure, dynamics, and function, re-structuring molecular biology.

Thus, while Feynman’s ‘jiggling and wiggling’ of complex molecules like proteins as random vibrations emphasizes their dynamics, it can misconstrue protein conformational changes. Proteins involve specific coordinated conformational changes that are essential for their specific functions to ensure life, such as enzyme catalysis or drug action. Modern research observes a complex and functionally relevant landscape of motion (Kern and Zuiderweg, Reference Kern and Zuiderweg2003; Otten et al., Reference Otten2018; van der Kamp et al., Reference van der Kamp2018; Wayment-Steele et al., Reference Wayment-Steele2024). These complex dynamics, occurring on various timescales, are crucial for understanding how macromolecules function and how signaling proceeds in the cell. The pioneering theory of the ‘energy landscape’ was proposed for protein folding (Frauenfelder et al., Reference Frauenfelder1991), providing a new framework for understanding how proteins spontaneously fold into their correct 3D structures. It shifted the paradigm from a rigid, step-by-step process to a more probabilistic, landscape-driven process, where the protein rapidly navigates a vast number of potential configurations to find the most stable one. In our current context, the landscape depiction led to the foundational functional implications, which described biological molecules pre-existing in many conformational states (dynamic ensembles), with external factors [e.g., binding partners, cellular signals like phosphorylation (Sinha et al., Reference Sinha2026)] shifting the population toward one of these states. This concept is now broadly recognized as the origin of molecular recognition, allostery, signaling, catalysis, and more (Nussinov, Reference Nussinov2025).

The concept of dynamic landscapes for function (Nussinov and Wolynes, Reference Nussinov and Wolynes2014), powered by physics and chemistry, precipitated the revolution of the dynamic structure–function paradigm that has come to dominate molecular biology. It provided understanding, resolving the anomalies that the old (rigid structure) paradigm could not explain, triggering the Kuhnian revolution. Dynamic conformational ensembles, as depicted by free energy landscapes, can make the molecular biology revolution work.

The understanding that biomacromolecules, like proteins and RNA, exist in many states that can be captured in free energy landscapes, has been a breakthrough (Zhou et al., Reference Zhou2001; Solomatin et al., Reference Solomatin2010; Pietrucci, Reference Pietrucci2017; Alhadeff et al., Reference Alhadeff2018; Sengupta and Rief, Reference Sengupta and Rief2021; Bose et al., Reference Bose2024; Degenhardt et al., Reference Degenhardt2025; Lee et al., Reference Lee2025; Zhang et al., Reference Zhang2025b). It clarified that biomacromolecules conform to the laws of physics and structural chemistry, and that they can be described in terms of the statistics of their conformations, which correspond to their energies (Shortle et al., Reference Shortle1998; Gavalda-Garcia et al., Reference Gavalda-Garcia2024). It implied that crystal structures may represent a conformation caught in the crystal lattice, not necessarily the one to target in drug design (Lyu et al., Reference Lyu2024; Hong et al., Reference Hong2025; Platzer et al., Reference Platzer2025). The coining of allostery by Monod and Jacob in 1961 (Monod and Jacob, Reference Monod and Jacob1961; Cui et al., Reference Cui2024), later applied to coupling between distant sites influencing function (Changeux, Reference Changeux2012; Cui et al., Reference Cui2024), has also fallen under this category. While allostery, a mechanism for regulating protein function by molecule binding to a site other than the active site, was a major advance in biochemistry, today we know that allostery results from the protein existing in an ensemble of conformations, as portrayed by the revolutionary concept of the free energy landscape – not two conformations as originally conceived (Popovych et al., Reference Popovych2006; Kar et al., Reference Kar2010; Guo and Zhou, Reference Guo and Zhou2016; Nussinov, Reference Nussinov2016; Wu et al., Reference Wu2024; Cui, Reference Cui2025).

Disordered proteins tend to have low barriers and shallow basins

The energy landscapes of disordered proteins suggest conformational continuums (Kulkarni et al., Reference Kulkarni2022), and multiple strategies have been developed to define, map, and use them [e.g., (de Aquino et al., Reference de Aquino2020; Luong et al., Reference Luong2022)]. Their landscapes exhibit multiple shallow minima basins separated by low energy barriers, allowing rapid interconversion among a wide range of conformations (Figure 3). They were described as multifunneled, with structures resembling the bound state being relatively unpopulated in equilibrium (Chebaro et al., Reference Chebaro2015). This could be explained by the structural entropy that would need to be overcome to reach competent states, which may cost sufficiently high free energy, making the probability of observing such conformations small, thus virtually invisible (Papoian, Reference Papoian2008). The landscape was viewed as organized around a certain state, which was characterized as a ‘weak funnel’ (Papoian and Wolynes, Reference Papoian and Wolynes2003), due to a weak driving force for folding. However, binding a target, often through favorable hydrophobic interactions, deepens the funnel, driving subsequent folding (Tsai et al., Reference Tsai2001; Papoian, Reference Papoian2008). The low energy barriers result from a lack of strong stabilizing hydrophobic interactions and a high frequency of (complementary) charged residues. The lack of a strong hydrophobic core implies that there is no single, deep energy minimum. The absence of localized strong stabilizing forces contributes to a ‘frustrated’ energy landscape where many states have similar energies (Mattos and Clark, Reference Mattos and Clark2008; Ferreiro et al., Reference Ferreiro2014; Gianni et al., Reference Gianni2021).

Figure 3.

The energy landscape of intrinsically disordered proteins (IDPs). Unlike globular proteins, which have a deep, narrow folding funnel that leads to a single, low-entropy native state, IDPs have a broad, relatively flat energy landscape (top panel). This landscape is characterized by numerous shallow energy minima and high configurational entropy. These features allow IDPs to exist as a dynamic ensemble of states rather than a single fixed structure. Many IDPs undergo induced folding, a process in which they transition from a disordered state to a structured one when they interact with a partner. This transition involves a significant loss of configurational entropy, a thermodynamically unfavorable change that is typically compensated by the favorable enthalpy generated through new intermolecular interactions. This ‘induced folding’ mechanism is a hallmark of IDP molecular recognition, enabling highly specific yet reversible signaling interactions. The interaction between POSH and Rac1 is a prime example of induced folding in IDPs (bottom panel). POSH (plenty of SH3s), also known as the E3 ubiquitin-protein ligase SH3RF1, acts as a scaffold protein that mediates JNK activation and apoptosis, or programmed cell death. POSH is a multidomain protein composed largely of intrinsically disordered regions; however, binding to active Rac1 triggers a structural transition. Specifically, the disordered POSH segment (residues 323–344) folds into a distinct α-helix–turn–β-strand conformation (PDB ID: 9RFB) upon complex formation.

The low energy barriers also allow intrinsically disordered proteins to rapidly transition between a wide variety of transient conformations, leading to faster allosteric communication. However, the multiple low barriers are also likely to reduce the allosteric accuracy. Since multiple pathways could be at play, the fidelity of the signal can be compromised. Allosteric communication involves shifts in conformational ensembles, and lower energy barriers between the states allow for rapid transitions, including on the microsecond to millisecond timescale, enabling the cell to adapt quickly to environmental changes. At the same time, the nature of multiple low-barrier transitions and pathways can diminish the specificity of signal transmission.

Disordered proteins have reduced structural content and higher conformational heterogeneity as compared to well-folded proteins (Uversky, Reference Uversky2016; Trivedi and Nagarajaram, Reference Trivedi and Nagarajaram2022). Despite this, AlphaFold was shown to be capable of modeling them, likely since the types of interactions between residues that stabilize the native states of disordered proteins do not fundamentally differ from those of folded proteins, although their context might (Brotzakis et al., Reference Brotzakis2025). Alternatively, this capability relates to training on the Protein Data Bank (PDB), which mostly contains stable, folded protein structures, but also bound states of intrinsically disordered regions. When bound to partners, the complexes of intrinsically disordered proteins often remain highly dynamic, i.e., fuzzy, with the complex itself residing in a shallow, yet broad minimum on the energy landscape (Fuxreiter, Reference Fuxreiter2020; Freiberger et al., Reference Freiberger2021). They are major contributors to the formation of liquid–liquid phase separation (LLPS), concentrating into liquid-like droplets (biomolecular condensates or membraneless organelles), where they tend to form multivalent interactions (Brocca et al., Reference Brocca2020; Nussinov and Jang Reference Nussinov and Jang2026; Nussinov et al., Reference Nussinov2026). As to posttranslational modifications (PTMs), which are common in disordered proteins, NMR spectroscopy and molecular dynamics simulations observed that acetylation reduces the conformational heterogeneity of the H4 tail, positioning the respective regions within the conformational space uniquely occupied by the conformation, which was proposed as correlating with specific biological functions (Viegas et al., Reference Viegas2025).

Collectively, lacking hydrophobic cores and with abundant electrostatic interactions, disordered proteins exhibit high fluctuations, constantly interconverting among states and microstates, a defining characteristic of their nature and function, captured by free energy landscapes. The PTM chemistry and location of the covalent attachments provide additional challenging complexities in modeling their ensembles.

Protein conformational transitions: The key to biological function

Proteins continuously sample their energy landscapes, always ‘hopping’ from one conformation to another. The interplay between their thermal (stochastic) fluctuations and their conformational changes is the cornerstone mechanism of protein function, binding, and cell signaling. While thermal fluctuations provide the energy for motion (making individual atomic movements stochastic), the transitions between functional states are guided by specific, predictable (governed by Boltzmann statistics and kinetic rates) pathways, even if a single molecule’s exact path at a given microsecond is random and governed by the shape of the landscape. Function is rooted in the probabilities of specific state transitions, which depend on the heights of the barriers between them and the relative stability of the respective conformations. Lower barriers may suggest a role in favored communication pathways in allostery. Higher barriers imply higher chances of trapping specific conformations. Those conformations may represent autoinhibited states and serve as targets for drug discovery. That speed is important, and it is not necessarily the case that faster is better, was shown by NMR dynamics of KaiB (a component of the circadian clock molecular machinery in cyanobacteria), which undergoes a significant conformational change to an alternate, functionally active state, to interact with KaiC. This transition involves a high energy barrier and occurs on a timescale of hours, matching the circadian rhythm period (Wayment-Steele et al., Reference Wayment-Steele2024). Processes can also be slowed down by reduced protein translation rates, resulting in exiting the cell cycle in differentiation or influenced by the rate of protein state conversion reminiscent of ribosome stalling by synonymous SNPs in P-glycoprotein 1 (Figure 2). As discussed above, a synonymous (rare) codon substitution can lead to altered folding pathways and distinct minima by fold-switching through kinetically trapped states (Tsai et al., Reference Tsai2008b). The long-enough ribosomal pause timescales may have resulted in altered structure and function of P-glycoprotein 1.

High energy barriers can hinder protein allosteric communication by trapping the protein in a specific conformational state. While the focus is often on proteins with efficient allostery, cascading down their pathways (Nussinov et al., Reference Nussinov2025a; Regev et al., Reference Regev2025), examples illustrating how high barriers impede communication include Epac2 activation (Selvaratnam et al., Reference Selvaratnam2012; Robichaux 3rd and Cheng, Reference Robichaux and Cheng2018; Zhang et al., Reference Zhang2025a). In the absence of an allosteric effector (cAMP), the protein appears locked due to high energy barriers involved in hinge-bending motion, which is further hindered by an ionic latch (Figure 4). The effector lowers the barrier, shifting the conformational equilibrium. Of special note, recently, the Kalodimos team discovered how Src family kinases execute processive phosphorylation, showing that their ensembles embrace a transient state between the active and inactive states (Gough and Kalodimos, Reference Gough and Kalodimos2024; Cui, Reference Cui2025). This intermediate state facilitates the rapid release of ADP, acting to ensure efficient catalytic turnover.

Figure 4.

Conformational switch of Epac2 induced by cAMP. Epac2, also known as RapGEF4, acts as a guanine nucleotide exchange factor (GEF) for small GTPases such as Rap1 and Rap2. Epac2 has multiple isoforms, including Epac2A, Epac2B, and Epac2C (top panel). The canonical form of Epac2 is Epac2A that contains two cyclic nucleotide-binding domains (CNBD-A and CNBD-B) and a Dishevelled, Egl-10, and Pleckstrin (DEP) domain in the N-terminal regulatory domain. The C-terminal catalytic domain contains a Ras exchange motif (REM), a Ras association (RA) domain, and a CDC25 homology domain (CDC25-HD or GEF domain). The CDC25-HD is responsible for activating Rap1. Epac2B lacks the CNBD-A similar to Epac1. Epac2C lacks both CNBD-A and DEP domain. The second messenger cyclic AMP (cAMP) is synthesized from ATP by adenylyl cyclase (AC) in response to the stimulation of Gαs-coupled G protein-coupled receptors (GPCRs) (bottom left panel). Once produced, cAMP binds to the regulatory domain of autoinhibited Epac2, inducing a conformational shift that releases its catalytic domain. This activation allows Epac2 to be recruited by Ras to the plasma membrane, facilitating nucleotide exchange on Rap1. Activated Rap1 then orchestrates a diverse signaling network. It recruits PI3K to promote cell growth and triggers the MAPK pathway to stimulate cell proliferation. Rap1 also engages effectors, such as RIAM, AF-6, and RapL, to regulate cell adhesion, migration, and polarity. Crystal structures of the full-length apo-Epac2 (PDB ID: 2byv, Mus musculus) in surface representation and cAMP-bound, active Epac2 (PDB ID: 3cf, Mus musculus) in cartoon representation (bottom right panel). In the active form, both the CNBD-A and the DEP domain are absent. Rap1, in surface representation, binds to the exposed catalytic core of the CDC25-HD. Part of graphics created with BioRender.

The free energy landscape provides a quantitative framework for understanding autoinhibition by illustrating the stability and kinetics of the protein’s different functional states (inactive versus active). Autoinhibition is characterized by the presence of a (relatively) deep free energy minimum corresponding to the inactive (autoinhibited) state, separated from the active state (also a minimum, but often higher in free energy in the absence of an activating signal) by a significant free energy barrier. The landscape shows energy barriers that must be overcome for the protein to switch between the inactive and active conformations. The height of these barriers dictates the transition rates; higher barriers mean slower activation/inactivation. Examples include protein tyrosine kinase Src, which is kept in an inactive, autoinhibited state when a phosphorylated tyrosine residue at its C-terminus (Tyr527) binds to its own SH2 domain (Levinson et al., Reference Levinson2008; Koudelkova et al., Reference Koudelkova2023); focal adhesion kinase (FAK) where in its resting state, the FERM domain of FAK folds back to bind to and block the catalytic cleft of its own kinase domain, preventing autophosphorylation and activation (Papusheva et al., Reference Papusheva2009; Brod and Fassler, Reference Brod and Fassler2020); myosin motor proteins (e.g., myosin II, V, and VI) employ autoinhibition in the absence of cargo to suppress their motor activity until a specific stimulus triggers activation and movement (Niu et al., Reference Niu2024); talin and vinculin adaptor proteins, which are crucial for cell adhesion, exist in compact, autoinhibited conformations where their ligand-binding sites for integrins and actin are masked by intramolecular head–tail interactions. Mechanical force can relieve this inhibition (Boujemaa-Paterski et al., Reference Boujemaa-Paterski2020; Bachmann et al., Reference Bachmann2023), and integrin cell surface adhesion receptors exist in a low-affinity, bent, and closed conformation. Binding of intracellular proteins like talin relieves the autoinhibition and shifts the integrin to an extended, high-affinity conformation (Pang et al., Reference Pang2023); calcium/calmodulin-dependent kinase II (CaMKII), which forms a dodecameric structure that is autoinhibited until the binding of calcium/calmodulin, which relieves the inhibition and allows the kinase to become active (Chao et al., Reference Chao2011); and Bruton’s tyrosine kinase (BTK), which is regulated by autoinhibition involving interactions between its domains, relieved upon B-cell activation through phosphorylation events (Wang et al., Reference Wang2015). Additional examples include c-Abl wild-type state (Liu et al., Reference Liu2022), B-Raf (Zhang et al., Reference Zhang2021), MEK (Maloney et al., Reference Maloney2022), PDK1 (Xu et al., Reference Xu2024), PTEN (Nussinov et al., Reference Nussinov2021), Ras (Dudas et al., Reference Dudas2020), and EGFR, where, before ligand binding, the dimerization domain of the extracellular region is blocked. Ligand binding releases this autoinhibition, allowing the receptor monomers to dimerize and become active. Allosteric PTMs also regulate function by promoting conformational transitions (Nussinov et al., Reference Nussinov2012; Sinha et al., Reference Sinha2026).

Conclusions and perspective

Physics and chemistry provide foundational principles upon which modern molecular biology is built, driving significant scientific advances and innovations. A landscape view applies not only to folding, which was its bedrock, but as demonstrated by the conformational selection and population shift paradigm, which is built upon it, to translational science too (Tsai and Nussinov, Reference Tsai and Nussinov2014; Nussinov et al., Reference Nussinov2025c). Different than folding funnels, translation centers on the heterogeneous ensembles around the funnel bottom. The conformational selection and population shift paradigm embraces mutational states made more populated by the population shift. Constitutive mutations can destabilize the inactive state, stabilize the active state, or both. The population shift is accomplished by allosteric networks and signaling bias (Nussinov et al., Reference Nussinov2025b) within this paradigm framework.

The groundbreaking concept that function is predicated on the statistics of populations, rather than one (or two) rigid structures manifested in crystals, and that biological processes are enacted by shifts of the conformational ensembles has been confirmed by numerous observations (Boehr et al., Reference Boehr2009). Considering the earlier prevailing belief that only one or two protein conformations are functional, the landscape view, which overturned it, is a (Kuhnian) revolution. Kinases serve as a prime example of this paradigm; their energy landscapes share conformational features across different kinase families, underscoring the crucial importance of landscapes portraying ensembles in deciphering function (Xie et al., Reference Xie2020; Gough and Kalodimos, Reference Gough and Kalodimos2024). Ultimately, the landscape framework clarifies the complex dynamics of enzyme states, facilitating drug discovery and design by identifying druggable target conformations.

Function (dysfunction) leverages dynamic transitions between pre-existing conformations, revolutionizing the understanding of allostery and, broadly, cellular regulation. Function can also be envisioned in the framework of evolution. The funnels concept, as conceived by Wolynes and his colleagues, resolved the Levinthal paradox by leveraging evolution. They harnessed the principle of minimal frustration, proposing that proteins are optimized by natural selection to minimize conflicts. In our current context, evolution plays a role through dynamic distributions of conformational ensembles, with the ensembles linking genotype to phenotype (Nussinov et al., Reference Nussinov2019).

The landscape concept has not only brought forth new theoretical interpretive tools, and inspired new experimental approaches, but bore clinical fruit (Nussinov and Wolynes, Reference Nussinov and Wolynes2014), leading to next-generation drug discovery (Nussinov et al., Reference Nussinov2025c).