The group has been active for a long time in the development of enhanced sampling methods. Tipically, such methods can be collective variable (CV) based, i.e. they accelerate specific physical degrees of freedom of the system, or CV-agnostic, like parallel tempering, where replicas of the system are sampled in parallel for different system parameters (e.g. temperature) to aid the exploration of the system's conformational space. However, the former methods usually struggle in intricate systems, such as large biophysical complexes, where the quality of the CV is hindered by the difficulty of encoding in it all the relevant degrees of freedom. Parallel techniques, instead, tend to become computationally very demanding and may still mostly explore less relevant high-energy portions of the system's phase space. Motivated by these shortcomes, the group developed OneOPES (Rizzi et al.), a replica exchange method that addresses the sampling of suboptimal CVs in complex systems by combining both a CV-based approach with the exploratory advantage of CV-agnostic techniques. Specifically, the main idea of OneOPES is that of simulating in parallel a handful of replicas, usually four to eight, where the lower replicas are more convergence oriented while the higher ones are more exploratory, i.e. they serve as sources of new states to seed the lower replicas and aid the exploration of the phase space.

OneOPES has been already applied successfully to a number of complex system both within and outside FLG's group, such as in ligand binding (Karrenbrock et al.), host-guest systems (Febrer Martinez et al.), GPCR activation (Aureli et al.), and ion channel modulation (Türkaydin et al.), among others, while the group also provides a GitHub tutorial covering its usage.

The prediction of ligand binding free energies is a fundamental task in drug discovery, as it can help identifying drug candidates and characterize their interaction with their targets. With respect to other computational techniques such as docking, MD simulations account for the dynamics and flexibility of both the ligand and the target, seldom times a protein, thus depicting an overall more realistic view of a host-ligand system. All atoms simulations can offer precious insights into the binding mode, its stability, the flexibility of the binding site and the ligand within it, the mechanism of binding, the main stable and transient host-ligand interactions, and the effect of specific mutations, among others.

Withing this framework, the group has been active in both the development of new methods and the validation of existing parametrizations. OneOPES has been successfully used to predict protein-ligand absolute binding free energies without the need to tailor the collective variables to each system (Karrenbrock et al.), and to map the complete binding and folding landscapes of Bcl-2 proteins, which are critical cell-death regulators in intrinsic apoptosis (Hanke et al.). Thanks to the robustness and transferability of the OneOPES binding protocol, FLG’s group has systematically investigated force fields, ligand parametrization, and water models for host-guest systems (Febrer Martinez et al.). The group also recently showed how simple solvation CVs, carefully crafted by analyzing the critical hydration sites in host-guest systems, can improve the convergence of binding free energies, underscoring the central role of water in molecular recognition (Schulze et al.).

G protein-coupled receptors¶

Kinases¶

Kinases are protein enzymes that catalyze the phosphorylation of substrates by transferring the phosphate moieties from donors, e.g. ATP and GTP, onto other downstream molecules like proteins, lipids, and small metabolites. The phosphorylation can alter the target’s activity, localization, stability, and interactions, thus making kinases fundamental regulators of many cell mechanisms, such as signaling, metabolism, and growth, among others. Consequently, they are of paramount importance for drug development, particularly in the context of cancer research.

FLG's group has been active for many years in the investigation of kinases. The team used molecular dynamics and enhanced sampling methods to provide seminal fundamental insights into kinases' conformational transitions (Berteotti et al.), into the activation mechanisms of adhesion kinases (Goñi et al.), and into the effect of oncogenic mutations on the conformational free-energy landscape (Sutto and Gervasio) and the ability to form oligomers (Sumanth Iyer et al., Galdadas et al.) of the epidermal grow factor receptor, among others. Lately, in a combined effort with experimental partners, the groups integrated cryo-electron microscopy, biophysical techniques, and molecular dynamics simulations to construct a model of the active complexes between MKK6 and p38α (Juyoux et al.) and MEK1 and ERK2 (von Velsen et al.), shedding light on their specific interactions, selectivity, and the overall mechanism of activation.

---

Apoptosis¶

The B cell lymphoma-2 (Bcl-2) protein family is the major regulator of intrinsic apoptosis. Its members control the permeabilization of the mitochondrial outer membrane (MOM), a hallmark process of programmed cell death. The family consists of three subgroups, namely (i) the pro-apoptotic effectors, or executioners, which oligomerize at the MOM and permeabilize it; (ii) the pro-survival anti-apoptotic members, which hinder the permeabilization of the MOM by binding to the pro-apoptotic members and inhibiting their activation; and (iii) the BH3-only initiators, which promote MOM permeabilization by activating pro-apoptotic executioners and/or by binding and antagonizing anti-apoptotic members. Archetypal examples for these subgroups are BAX, Bcl-XL, and BID, respectively. By finely orchestrating pro- and anti-apoptotic signals, the Bcl-2 proteins regulate cell survival, thus making them important therapeutic targets for cancer and other complex diseases to pharmaceutically regulate apoptosis. Although the interaction network of the Bcl-2 family members has become clearer and better defined in recent years, a complete and unambiguous picture of the intrinsic apoptotic pathway is still lacking.

To explore and characterize the Bcl-2 apoptotic pathway, FLG’s group combines enhanced sampling with unbiased all-atoms and coarse grained MD simulations to complement with atomistic insights the experimental data of partner scientists at the University of Geneva. Notably, the group recently characterized the structural ensemble of the inhibitory Bcl-xL/tBid complex at the mitochondrial membrane, showing that Bcl-xL and tBid form a heterodimer anchored to the membrane by the C-terminal helix of Bcl-xL (Elsner et al.), and investigated the binding of Bid peptides with full-length membrane-anchored Bcl-xL and Bax, identifying shared and unique features as a function of their distinct apoptotic roles (Hanke et al.).

---

Cryptic pockets¶

Cryptic pockets are hidden protein pockets that open because of conformational changes or protein dynamics. Since these pockets are generally not accessible, they do not constitute visible cavities and are thus not resolved in experiments such as X-ray crystallography or cryo-EM imaging, making them extremely hard to find and characterize. Cryptic pockets can emerge via thermal fluctuation, because of a conformational shift in the protein, or following the approach of a ligand. Notably, ligands can both trigger a cavity opening or trap a transient state of the protein, mechanisms known as induced-fit and conformational selection, respectively. Since the cavities are short-lived and hidden at rest, and the processes underpinning their opening are very complex and partially unknown, cryptic pockets are notoriusly hard to find. At the same time, however, they are considered of extreme interest for drug development as they may offer new targets for proteins that lack a traditional accessible active site, effectively making "druggable" thousands of proteins that are considered "undruggable", that is, impossible to target with traditional small-molecule drugs.

FLG's group has been working on MD methods to investigate the dynamics and systematically detect druggable cryptic pockets in targets of biopharmaceutical interest. In this context, the group developed SWISH (Oleinikovas et al.) and SWISH-X (Borsatto et al.), two enhanced sampling algorithms based on Hamiltonian replica exchange that enhance the conformational sampling of proteins by scaling the interactions of apolar atoms with water and stabilize the pockets with mixed-solvents or organic fragments. In SWISH-X the replicas are also distributed along a temperature ramp to further aid close-to-open transitions. SWISH succeeded in exploring the nontrivial cryptic binding site in Nsp1, an important SARS-CoV-2 target (Borsatto et al.). The group is also tackling the crypting pockets discovery problem by leveraging modern AI tools. Specifically, the team built a large-scale database of cryptic sites, identified by applying a machine learning model to detect ligand-induced conformational changes, and revealed that cryptic pockets are widespread, occurring in ~18% of protein clusters (Febrer-Martinez et al.). The database resource was then used to fine-tuned a protein language model to predict cryptic sites directly from protein sequences. The model is implemented in CryptoBank, a publicly available web server.