G-quadruplex-binding proteins (G4BPs) play key roles in RNA metabolism and stress response, yet their identification remains experimentally challenging. Here, we present a deep learning (DL) framework for the prediction of RNA G4BPs (RG4BPs), integrating diverse encoding strategies and neural architectures. Our best-performing model, which includes ESM-2 protein language model embeddings and consists of an LSTM architecture, achieved 86% accuracy in distinguishing RG4BPs from non-binder proteins. The application of this model to the human proteome uncovered 2160 high-confidence RG4BP candidates, many of which display intrinsically disordered regions (IDRs) and enrichment in stress granule organelles. These findings reveal a potential link between G-quadruplex recognition and cellular stress responses. To enable easy and broad access to the framework, we developed G⁴REP, a web server for RG4BP prediction and analysis. Overall, an effective approach to explore the RG4BPs landscape and uncover novel players in RNA regulation is provided.

The CASP16 experiment provided the first opportunity to benchmark AlphaFold3. In contrast to AlphaFold2, AlphaFold3 can predict the structure of non-protein molecules. According to the benchmark presented by the developers, it is expected to perform slightly better than AlphaFold2 for proteins. In this study, we assess the performance of AlphaFold3 using both automatic server submissions (AF3-server) and manual predictions from the Elofsson group (Elofsson). All predictions were generated via the AlphaFold3 web server, with manual interventions applied to large targets and ligands. Compared to AlphaFold2-based methods, we found that AlphaFold3 performs slightly better for protein complexes. However, when massive sampling is applied to AlphaFold2, the difference disappears. It was also noted that, according to the official ranking from CASP, the AF3-server performs better than AlphaFold2 for easier targets, but not for harder targets. Furthermore, the performance of the AF3-server is comparable to the best methods when considering the top-ranked predictions, but slightly behind when examining the best among the five submitted models. Here, there exist targets where AF3-server, the top-ranked method, is worse than lower-ranked models, indicating that a venue for progress could be to develop better strategies for identifying the best out of the generated models. When using AF3-server to predict the stoichiometry of larger protein complexes, the accuracy is limited, especially for heteromeric targets. When analyzing the predictions including nucleic acids, it was found that, in general, the accuracy is relatively low. However, the AF3-server performance was not far behind that of the top-ranked method. In summary, AF3-server offers a user-friendly tool that provides predictions comparable to state-of-the-art methods in all categories of CASP.

The Critical Assessment of Structure Prediction (CASP) brings together a diverse group of scientists, from deep learning experts to NMR specialists, all aimed at developing accurate prediction algorithms that can effectively characterize the structural aspects of biomolecules relevant to their functions. Engagement within the CASP community has traditionally been limited to the prediction season and the conference, with limited discourse in the 1.5 years between CASP seasons. CASP special interest groups (SIGs) were established in 2023 to encourage continuous dialogue within the community. The online seminar series has drawn global participation from across disciplines and career stages. This has facilitated cross-disciplinary discussions fostering collaborations. The archives of these seminars have become a vital learning tool for newcomers to the field, lowering the barrier to entry.

Motivation: AlphaFold2 significantly improved the prediction of protein complex structures. However, its accuracy is lower for interactions without coevolutionary signals, such as host–pathogen and antibody–antigen interactions. Two strategies have been developed to address this limitation: massive sampling and replacing the evoformer with the pairformer, which does not rely on coevolution, as introduced in AlphaFold3, thereby enabling more structural reasoning by the network.

Results: In this study, we benchmark structure prediction methods on unseen antibody–antigen complexes. We found that increased sampling improves the chances of generating a correct protein model, roughly in a log-linear manner. However, the internal quality estimates by AlphaFold often cannot identify the best predicted structures for each target, resulting in a significant loss of performance for the top-ranked protein model compared with the best model. For all methods, a significant challenge remains the identification of the best model. We also show that AlphaFold3 outperforms AlphaFold2, Boltz-1, and Chai-1. Furthermore, AlphaFold3 performance declines significantly for complexes that lack structural similarity to the training set, indicating that it has to some extent learned to detect remote structural similarities.

Availability and implementation: All code is available from github.com/samuelfromm/abag-benchmark-set/ and all data from DOI: 10.5281/zenodo.17978681. The latter repository also contains the code.

Molecular structure generation is a fundamental problem that involves determining the 3D positions of molecules’ constituents. It has crucial biological applications, such as molecular docking, protein folding, and molecular design. Recent advances in generative modeling, such as diffusion models and flow matching, have made great progress on these tasks by modeling molecular conformations as a distribution. In this work, we focus on flow matching and adopt an energy-based perspective to improve training and inference of structure generation models. Our view results in a mapping function, represented by a deep network, that is directly learned to iteratively map random configurations, i.e. samples from the source distribution, to target structures, i.e. points in the data manifold. This yields a conceptually simple and empirically effective flow matching setup that is theoretically justified and has interesting connections to fundamental properties such as idempotency and stability, as well as the empirically useful techniques such as structure refinement in AlphaFold. Experiments on protein docking as well as protein backbone generation consistently demonstrate the method’s effectiveness, where it outperforms recent baselines of task-associated flow matching and diffusion models, using a similar computational budget.

Molecular structure generation is a fundamental problem that involves determining the 3D positions of molecules’ constituents. It has crucial biological applications, such as molecular docking, protein folding, and molecular design. Recent advances in generative modeling, such as diffusion models and flow matching, have made great progress on these tasks by modeling molecular conformations as a distribution. In this work, we focus on flow matching and adopt an energy-based perspective to improve training and inference of structure generation models. Our view results in a mapping function, represented by a deep network, that is directly learned to iteratively map random configurations, i.e. samples from the source distribution, to target structures, i.e. points in the data manifold. This yields a conceptually simple and empirically effective flow matching setup that is theoretically justified and has interesting connections to fundamental properties such as idempotency and stability, as well as the empirically useful techniques such as structure refinement in AlphaFold. Experiments on protein docking as well as protein backbone generation consistently demonstrate the method’s effectiveness, where it outperforms recent baselines of task-associated flow matching and diffusion models, using a similar computational budget.

Recent advances in geometric deep learning and generative modeling have enabled the design of novel proteins with a wide range of desired properties. However, current state-of-the-art approaches are typically restricted to generating proteins with only static target properties, such as motifs and symmetries. In this work, we take a step towards overcoming this limitation by proposing a framework to condition structure generation on flexibility, which is crucial for key functionalities such as catalysis or molecular recognition. We first introduce BackFlip, an equivariant neural network for predicting per-residue flexibility from an input backbone structure. Relying on BackFlip, we propose FliPS, an SE(3)-equivariant conditional flow matching model that solves the inverse problem, that is, generating backbones that display a target flexibility profile. In our experiments, we show that FliPS is able to generate novel and diverse protein backbones with the desired flexibility, verified by Molecular Dynamics (MD) simulations. FliPS and BackFlip are available at https://github.com/graeter-group/flips.

Recent advances in geometric deep learning and generative modeling have enabled the design of novel proteins with a wide range of desired properties. However, current state-of-the-art approaches are typically restricted to generating proteins with only static target properties, such as motifs and symmetries. In this work, we take a step towards overcoming this limitation by proposing a framework to condition structure generation on flexibility, which is crucial for key functionalities such as catalysis or molecular recognition. We first introduce BackFlip, an equivariant neural network for predicting per-residue flexibility from an input backbone structure. Relying on BackFlip, we propose FliPS, an SE(3)-equivariant conditional flow matching model that solves the inverse problem, that is, generating backbones that display a target flexibility profile. In our experiments, we show that FliPS is able to generate novel and diverse protein backbones with the desired flexibility, verified by Molecular Dynamics (MD) simulations. FliPS and BackFlip are available at https://github.com/graeter-group/flips.

Disordered regions are an important functional feature of many multidomain proteins. A prime example is proteins in membraneless organelles, which contain folded domains that engage in specific interactions and disordered low-complexity (LC) domains that mediate liquid-liquid phase separation. Studying these complex architectures remains challenging due to their conformational variability. Native mass spectrometry (nMS) is routinely employed to analyze conformations and interactions of folded or disordered proteins; however, its ability to analyze proteins with disordered LC domains has not been investigated. Here, we analyze the ionization and conformational states of designed model proteins that recapitulate key features of proteins found in membraneless organelles. Our results show that charge state distributions (CSDs) in nMS reflect partial disorder regardless of the protein sequence, providing insights into their conformational plasticity and interactions. By applying the same CSD analysis to a spider silk protein fragment, we find that interactions between folded domains that trigger silk assembly simultaneously induce conformational changes in the LC domains. Lastly, using intact nucleosomes, we demonstrate that CSDs are a good predictor for the disorder content of complex native assemblies. We conclude that nMS reliably informs about the conformational landscape of proteins with LC domains, which is crucial for understanding protein condensates in cellular environments.