共找到 20 条结果
Protein language models are trained on highly imbalanced datasets, raising the question of how they represent underrepresented biological sequences. Using viral proteins as a case study across ESM model families, we identify a dominant nativeness axis in embedding space, aligned with masked reconstruction perplexity, that orders sequences from well-modeled cellular proteins through viral proteins to shuffled and random sequences. Scaling contracts this axis unevenly across viral families. Despite this, protein language model embeddings retain viral-specific signal: viral proteins remain linearly separable beyond zero-shot perplexity and shallow sequence features. Together, these results suggest that pLM representations are structured by a general notion of nativeness while preserving information specific to distinct biological groups.
We present reproducible, edge-aware baselines for ogbn-proteins in PyTorch Geometric (PyG). We study two system choices that dominate practice: (i) how 8-dimensional edge evidence is aggregated into node inputs, and (ii) how edges are used inside message passing. Our strongest baseline is GraphSAGE with sum-based edge-to-node features. We compare LayerNorm (LN), BatchNorm (BN), and a species-aware Conditional LayerNorm (CLN), and report compute cost (time, VRAM, parameters) together with accuracy (ROC-AUC) and decision quality. In our primary experimental setup (hidden size 512, 3 layers, 3 seeds), sum consistently beats mean and max; BN attains the best AUC, while CLN matches the AUC frontier with better thresholded F1. Finally, post-hoc per-label temperature scaling plus per-label thresholds substantially improves micro-F1 and expected calibration error (ECE) with negligible AUC change, and light label-correlation smoothing yields small additional gains. We release standardized artifacts and scripts used for all of the runs presented in the paper.
Proteins must bind to specific other proteins in vivo in order to function. The proteins must bind only to one or a few other proteins of the of order a thousand proteins typically present in vivo. Using a simple model of a protein, specific binding in many component mixtures is studied. It is found to be a demanding function in the sense that it demands that the binding sites of the proteins be encoded by long sequences of bits, and the requirement for specific binding then strongly constrains these sequences. This is quantified by the capacity of proteins of a given size (sequence length), which is the maximum number of specific-binding interactions possible in a mixture. This calculation of the maximum number possible is in the same spirit as the work of Shannon and others on the maximum rate of communication through noisy channels.
The length distribution of proteins measured in amino acids follows the CoHSI (Conservation of Hartley-Shannon Information) probability distribution. In previous papers we have verified various predictions of this using the Uniprot database but here we explore a novel predicted relationship between the longest proteins and evolutionary time. We demonstrate from both theory and experiment that the longest protein and the total number of proteins are intimately related by Information Theory and we give a simple formula for this. We stress that no evolutionary explanation is necessary; it is an intrinsic property of a CoHSI system. While the CoHSI distribution favors the appearance of proteins with fewer than 750 amino acids (characteristic of most functional proteins or their constituent domains) its intrinsic asymptotic power-law also favors the appearance of unusually long proteins; we predict that there are as yet undiscovered proteins longer than 45,000 amino acids. In so doing, we draw an analogy between the process of protein folding driven by favorable pathways (or funnels) through the energy landscape of protein conformations, and the preferential information pathways through
Topological analysis of large scale protein-protein interaction networks (PINs) is important for understanding the organisational and functional principles of individual proteins. The number of interactions that a protein has in a PIN has been observed to be correlated with its indispensability. Essential proteins generally have more interactions than the non-essential ones. We show here that the lethality associated with removal of a protein from the yeast proteome correlates with different centrality measures of the nodes in the PIN, such as the closeness of a protein to many other proteins, or the number of pairs of proteins which need a specific protein as an intermediary in their communications, or the participation of a protein in different protein clusters in the PIN. These measures are significantly better than random selection in identifying essential proteins in a PIN. Centrality measures based on graph spectral properties of the network, in particular the subgraph centrality, show the best performance in identifying essential proteins in the yeast PIN. Subgraph centrality gives important structural information about the role of individual proteins and permits the selecti
DNA-binding proteins are a class of proteins which have a specific or general affinity to DNA and include three important components: transcription factors; nucleases, and histones. DNA-binding proteins also perform important roles in many types of cellular activities. In this paper we describe machine learning systems for the prediction of DNA- binding proteins where a Support Vector Machine and a Cascade Correlation Neural Network are optimized and then compared to determine the learning algorithm that achieves the best prediction performance. The information used for classification is derived from characteristics that include overall charge, patch size and amino acids composition. In total 121 DNA- binding proteins and 238 non-binding proteins are used to build and evaluate the system. For SVM using the ANOVA Kernel with Jack-knife evaluation, an accuracy of 86.7% has been achieved with 91.1% for sensitivity and 85.3% for specificity. For CCNN optimized over the entire dataset with Jack knife evaluation we report an accuracy of 75.4%, while the values of specificity and sensitivity achieved were 72.3% and 82.6%, respectively.
Intrinsically disordered proteins (IDPs) and multidomain proteins with flexible linkers show a high level of structural heterogeneity and are best described by ensembles consisting of multiple conformations with associated thermodynamic weights. Determining conformational ensembles usually involves integration of biophysical experiments and computational models. In this review, we discuss current approaches to determining conformational ensembles of IDPs and multidomain proteins, including the choice of biophysical experiments, computational models used to sample protein conformations, models to calculate experimental observables from protein structure, and methods to refine ensembles against experimental data. We also provide examples of recent applications of integrative conformational ensemble determination to study IDPs and multidomain proteins and suggest future directions for research in the field.
Protein identification is one of the major task of Proteomics researchers. Protein identification could be resumed by searching the best match between an experimental mass spectrum and proteins from a database. Nevertheless this approach can not be used to identify new proteins or protein variants. In this paper an evolutionary approach is proposed to discover new proteins or protein variants thanks a "de novo sequencing" method. This approach has been experimented on a specific grid called Grid5000 with simulated spectra and also real spectra.
We present an analysis of the effects of global topology on the structural stability of folded proteins in thermal equilibrium with a heat bath. For a large class of single domain proteins, we computed the harmonic spectrum within the Gaussian Network Model (GNM) and determined the spectral dimension, a parameter describing the low frequency behaviour of the density of modes. We find a surprisingly strong correlation between the spectral dimension and the number of amino acids of the protein. Considering that larger spectral dimension value relate to more topologically compact folded state, our results indicate that for a given temperature and length of the protein, the folded structure corresponds to the less compact folding compatible with thermodynamic stability.
Scaling of folding times in Go models of proteins and of decoy structures with the Lennard-Jones potentials in the native contacts reveal %robust power law trends when studied under optimal folding conditions. The power law exponent depends on the type of native geometry. Its value indicates lack of kinetic optimality in the model proteins. In proteins, mechanical and thermodynamic stabilities are correlated.
Spatially proximate amino acids in a protein tend to coevolve. A protein's three-dimensional (3D) structure hence leaves an echo of correlations in the evolutionary record. Reverse engineering 3D structures from such correlations is an open problem in structural biology, pursued with increasing vigor as more and more protein sequences continue to fill the data banks. Within this task lies a statistical inference problem, rooted in the following: correlation between two sites in a protein sequence can arise from firsthand interaction but can also be network-propagated via intermediate sites; observed correlation is not enough to guarantee proximity. To separate direct from indirect interactions is an instance of the general problem of inverse statistical mechanics, where the task is to learn model parameters (fields, couplings) from observables (magnetizations, correlations, samples) in large systems. In the context of protein sequences, the approach has been referred to as direct-coupling analysis. Here we show that the pseudolikelihood method, applied to 21-state Potts models describing the statistical properties of families of evolutionarily related proteins, significantly outper
A novel scheme is introduced to capture the spatial correlations of consecutive amino acids in naturally occurring proteins. This knowledge-based strategy is able to carry out optimally automated subdivisions of protein fragments into classes of similarity. The goal is to provide the minimal set of protein oligomers (termed ``oligons'' for brevity) that is able to represent any other fragment. At variance with previous studies where recurrent local motifs were classified, our concern is to provide simplified protein representations that have been optimised for use in automated folding and/or design attempts. In such contexts it is paramount to limit the number of degrees of freedom per amino acid without incurring in loss of accuracy of structural representations. The suggested method finds, by construction, the optimal compromise between these needs. Several possible oligon lengths are considered. It is shown that meaningful classifications cannot be done for lengths greater than 6 or smaller than 4. Different contexts are considered were oligons of length 5 or 6 are recommendable. With only a few dozen of oligons of such length, virtually any protein can be reproduced within typi
We use a three dimensional cubic lattice model of proteins to study their properties that determine folding to the native state. The protein chain is modeled as a sequence of $N$ beads. The interactions between beads are taken from a Gaussian distribution of energies. We studied 56 sequences with unique ground states for $N = 15$ and $27$. Thermodynamic and kinetic properties were determined using Monte Carlo simulations and exhaustive enumeration. For all sequences we find collapse temperature, $T_θ$, at which the protein collapses into compact structure, and folding temperature, $T_{f}$, at which the protein acquires the native state. We show that parameter $σ= (T_θ - T_{f})/T_θ$ correlates extremely well with folding times. Fast folders reach the native state via a nucleation collapse mechanism without forming any intermediates, whereas for moderate and slow folders only a fraction of molecules $Φ$ reaches the native state by this process. The remaining fraction folds via three stage multipathway process. The simultaneous requirement of native state stability and kinetic accessibility can be achieved for the sequences with small values of $σ$.
A technique has been developed for the separation of proteins by two-dimensional polyacrylamide gel electrophoresis. Due to its resolution and sensitivity, this technique is a powerful tool for the analysis and detection of proteins from complex biological sources. Proteins are separated according to isoelectric point by isoelectric focusing in the first dimension, and according to molecular weight by sodium dodecyl sulfate electrophoresis in the second dimension. Since these two parameters are unrelated, it is possible to obtain an almost uniform distribution of protein spots across a two-diminsional gel. This technique has resolved 1100 different components from Escherichia coli and should be capable of resolving a maximum of 5000 proteins. A protein containing as little as one disintegration per min of either 14C or 35S can be detected by autoradiography. A protein which constitutes 10 minus 4 to 10 minus 5% of the total protein can be detected and quantified by autoradiography. The reproducibility of the separation is sufficient to permit each spot on one separation to be matched with a spot on a different separation. This technique provides a method for estimation (at the described sensitivities) of the number of proteins made by any biological system. This system can resolve proteins differing in a single charge and consequently can be used in the analysis of in vivo modifications resulting in a change in charge. Proteins whose charge is changed by missense mutations can be identified. A detailed description of the methods as well as the characteristics of this system are presented.
Protein design is a fundamental challenge in biotechnology, aiming to design novel sequences with specific functions within the vast space of possible proteins. Recent advances in deep generative models have enabled function-based protein design from textual descriptions, yet struggle with structural plausibility. Inspired by classical protein design methods that leverage natural protein structures, we explore whether incorporating fragments from natural proteins can enhance foldability in generative models. Our empirical results show that even random incorporation of fragments improves foldability. Building on this insight, we introduce ProDVa, a novel protein design approach that integrates a text encoder for functional descriptions, a protein language model for designing proteins, and a fragment encoder to dynamically retrieve protein fragments based on textual functional descriptions. Experimental results demonstrate that our approach effectively designs protein sequences that are both functionally aligned and structurally plausible. Compared to state-of-the-art models, ProDVa achieves comparable function alignment using less than 0.04% of the training data, while designing sig
The prediction of protein-protein interactions (PPIs) is crucial for understanding biological functions and diseases. Previous machine learning approaches to PPI prediction mainly focus on direct physical interactions, ignoring the broader context of nonphysical connections through intermediate proteins, thus limiting their effectiveness. The emergence of Large Language Models (LLMs) provides a new opportunity for addressing this complex biological challenge. By transforming structured data into natural language prompts, we can map the relationships between proteins into texts. This approach allows LLMs to identify indirect connections between proteins, tracing the path from upstream to downstream. Therefore, we propose a novel framework ProLLM that employs an LLM tailored for PPI for the first time. Specifically, we propose Protein Chain of Thought (ProCoT), which replicates the biological mechanism of signaling pathways as natural language prompts. ProCoT considers a signaling pathway as a protein reasoning process, which starts from upstream proteins and passes through several intermediate proteins to transmit biological signals to downstream proteins. Thus, we can use ProCoT to
We demonstrate that Protein-Protein Interaction (PPI) networks in several eucaryotic organisms contain significantly more self-interacting proteins than expected if such homodimers randomly appeared in the course of the evolution. We also show that on average homodimers have twice as many interaction partners than non-self-interacting proteins. More specifically the likelihood of a protein to physically interact with itself was found to be proportional to the total number of its binding partners. These properties of dimers are are in agreement with a phenomenological model in which individual proteins differ from each other by the degree of their ``stickiness'' or general propensity towards interaction with other proteins including oneself. A duplication of self-interacting proteins creates a pair of paralogous proteins interacting with each other. We show that such pairs occur more frequently than could be explained by pure chance alone. Similar to homodimers, proteins involved in heterodimers with their paralogs on average have twice as many interacting partners than the rest of the network. The likelihood of a pair of paralogous proteins to interact with each other was also show
Strong excitonic coupling and photon antibunching (AB) have been observed together in Venus yellow fluorescent protein dimers and currently lack a cohesive theoretical explanation. In 2019, Kim et al. demonstrated Davydov splitting in circular dichroism spectra, revealing strong J-like coupling, while antibunched fluorescence emission was confirmed by combined antibunching--fluorescence correlation spectroscopy (AB/FCS fingerprinting). To investigate the implications of this coexistence, Venus yellow fluorescent protein (YFP) dimer population dynamics are modeled within a Lindblad master equation framework, testing its ability to cope with typical, data-informed, Venus YFP dimer time and energy values. Simulations predict multiple-femtosecond (fs) decoherence, yielding bright/dark state mixtures consistent with antibunched fluorescence emission at room temperature. Thus, excitonic coupling and photon AB in Venus YFP dimers are reconciled without invoking long-lived quantum coherence. However, clear violations of several Lindblad approximation validity conditions appear imminent, calling for careful modifications to choices of standard system and bath definitions and parameter value
Proteins are large biomolecules that regulate all living organisms and consist of one or several chains. The primary structure of a protein chain is a sequence of amino acid residues whose three main atoms (alpha-carbon, nitrogen, and carbonyl carbon) form a protein backbone. The tertiary structure is the rigid shape of a protein chain represented by atomic positions in 3-dimensional space. Because different geometric structures often have distinct functional properties, it is important to continuously quantify differences in rigid shapes of protein backbones. Unfortunately, many widely used similarities of proteins fail axioms of a distance metric and discontinuously change under tiny perturbations of atoms. This paper develops a complete invariant that identifies any protein backbone in 3-dimensional space, uniquely under rigid motion. This invariant is Lipschitz bi-continuous in the sense that it changes up to a constant multiple of a maximum perturbation of atoms, and vice versa. The new invariant has been used to detect thousands of (near-)duplicates in the Protein Data Bank, whose presence inevitably skews machine learning predictions. The resulting invariant space allows low
Membrane proteins typically deform the surrounding lipid bilayer membrane, which can play an important role in the function, regulation, and organization of membrane proteins. Membrane elasticity theory provides a beautiful description of protein-induced lipid bilayer deformations, in which all physical parameters can be directly determined from experiments. Analytic treatments of the membrane elasticity theory of protein-induced lipid bilayer deformations have largely focused on idealized protein shapes with circular cross section, and on perturbative solutions for proteins with non-circular cross section. We develop here a boundary value method (BVM) that permits the construction of non-perturbative analytic solutions of protein-induced lipid bilayer deformations for non-circular protein cross sections, for constant as well as variable boundary conditions along the bilayer-protein interface. We apply this BVM to protein-induced lipid bilayer thickness deformations. Our BVM reproduces available analytic solutions for proteins with circular cross section and yields, for proteins with non-circular cross section, excellent agreement with numerical, finite element solutions. On this b