ProTstab - predictor for cellular protein stability

TemBERTure: advancing protein thermostability prediction with deep learning and attention mechanisms

Motivation

Understanding protein thermostability is essential for numerous biotechnological applications, but traditional experimental methods are time-consuming, expensive, and error-prone. Recently, deep learning (DL) techniques from natural language processing (NLP) was extended to the field of biology, since the primary sequence of proteins can be viewed as a string of amino acids that follow a physicochemical grammar.

Results

In this study, we developed TemBERTure, a DL framework that predicts thermostability class and melting temperature from protein sequences. Our findings emphasize the importance of data diversity for training robust models, especially by including sequences from a wider range of organisms. Additionally, we suggest using attention scores from Deep Learning models to gain deeper insights into protein thermostability. Analyzing these scores in conjunction with the 3D protein structure can enhance understanding of the complex interactions among amino acid properties, their positioning, and the surrounding microenvironment. By addressing the limitations of current prediction methods and introducing new exploration avenues, this research paves the way for more accurate and informative protein thermostability predictions, ultimately accelerating advancements in protein engineering.

Availability and implementation

TemBERTure model and the data are available at: https://github.com/ibmm-unibe-ch/TemBERTure.

Review of predicting protein stability changes upon variations

Forecasting alterations in protein stability caused by variations holds immense importance. Improving the thermal stability of proteins is important for biomedical and industrial applications. This review discusses the latest methods for predicting the effects of mutations on protein stability, databases containing protein mutations and thermodynamic parameters, and experimental techniques for efficiently assessing protein stability in high-throughput settings. Various publicly available databases for protein stability prediction are introduced. Furthermore, state-of-the-art computational approaches for anticipating protein stability changes due to variants are reviewed. Each method's types of features, base algorithm, and prediction results are also detailed. Additionally, some experimental approaches for verifying the prediction results of computational methods are introduced. Finally, the review summarizes the progress and challenges of protein stability prediction and discusses potential models for future research directions.

Influence of Model Structures on Predictors of Protein Stability Changes from Single-Point Mutations

Missense variation in genomes can affect protein structure stability and, in turn, the cell physiology behavior. Predicting the impact of those variations is relevant, and the best-performing computational tools exploit the protein structure information. However, most of the current protein sequence variants are unresolved, and comparative or ab initio tools can provide a structure. Here, we evaluate the impact of model structures, compared to experimental structures, on the predictors of protein stability changes upon single-point mutations, where no significant changes are expected between the original and the mutated structures. We show that there are substantial differences among the computational tools. Methods that rely on coarse-grained representation are less sensitive to the underlying protein structures. In contrast, tools that exploit more detailed molecular representations are sensible to structures generated from comparative modeling, even on single-residue substitutions.

ProS-GNN: Predicting effects of mutations on protein stability using graph neural networks

Predicting protein stability change upon variation through a computational approach is a valuable tool to unveil the mechanisms of mutation-induced drug failure and develop immunotherapy strategies. Some previous machine learning-based techniques exhibit anti-symmetric bias toward destabilizing situations, whereas others struggle with generalization to unseen examples. To address these issues, we propose a gated graph neural network-based approach to predict changes in protein stability upon mutation. The model uses message passing to encode the links between the molecular structure and property after eliminating the non-mutant structure and creating input feature vectors. While doing so, it also incorporates the coordinates of the raw atoms to provide spatial insights into the chemical systems. We test the model on the Ssym, Myoglobin, Broom, and p53 datasets to demonstrate the generalization performance. Compared to existing approaches, our proposed method achieves improved linearity with symmetry in less time. The code for this study is available at: https://github.com/HongzhouTang/Pros-GNN.

PON-Fold: Prediction of Substitutions Affecting Protein Folding Rate

Most proteins fold into characteristic three-dimensional structures. The rate of folding and unfolding varies widely and can be affected by variations in proteins. We developed a novel machine-learning-based method for the prediction of the folding rate effects of amino acid substitutions in two-state folding proteins. We collected a data set of experimentally defined folding rates for variants and used them to train a gradient boosting algorithm starting with 1161 features. Two predictors were designed. The three-class classifier had, in blind tests, specificity and sensitivity ranging from 0.324 to 0.419 and from 0.256 to 0.451, respectively. The other tool was a regression predictor that showed a Pearson correlation coefficient of 0.525. The error measures, mean absolute error and mean squared error, were 0.581 and 0.603, respectively. One of the previously presented tools could be used for comparison with the blind test data set, our method called PON-Fold showed superior performance on all used measures. The applicability of the tool was tested by predicting all possible substitutions in a protein domain. Predictions for different conformations of proteins, open and closed forms of a protein kinase, and apo and holo forms of an enzyme indicated that the choice of the structure had a large impact on the outcome. PON-Fold is freely available.

Predicting thermostability difference between cellular protein orthologs

MOTIVATION

Protein thermostability is of great interest, both in theory and in practice.

RESULTS

This study compared orthologous proteins with different cellular thermostability. A large number of physicochemical properties of protein were calculated and used to develop a series of machine learning models for predicting cellular thermostability differences between orthologous proteins. Most of the important features in these models are also highly correlated to relative cellular thermostability. A comparison between the present study with previous comparison of orthologous proteins from thermophilic and mesophilic organisms found that most highly correlated features are consistent in these studies, suggesting they may be important to protein thermostability.

AVAILABILITY AND IMPLEMENTATION

Data freely available for download at https://github.com/fangj3/cellular-protein-thermostability-dataset.

DeepSTABp: A Deep Learning Approach for the Prediction of Thermal Protein Stability

Proteins are essential macromolecules that carry out a plethora of biological functions. The thermal stability of proteins is an important property that affects their function and determines their suitability for various applications. However, current experimental approaches, primarily thermal proteome profiling, are expensive, labor-intensive, and have limited proteome and species coverage. To close the gap between available experimental data and sequence information, a novel protein thermal stability predictor called DeepSTABp has been developed. DeepSTABp uses a transformer-based protein language model for sequence embedding and state-of-the-art feature extraction in combination with other deep learning techniques for end-to-end protein melting temperature prediction. DeepSTABp can predict the thermal stability of a wide range of proteins, making it a powerful and efficient tool for large-scale prediction. The model captures the structural and biological properties that impact protein stability, and it allows for the identification of the structural features that contribute to protein stability. DeepSTABp is available to the public via a user-friendly web interface, making it accessible to researchers in various fields.

DeepTP: A Deep Learning Model for Thermophilic Protein Prediction

Thermophilic proteins have important value in the fields of biopharmaceuticals and enzyme engineering. Most existing thermophilic protein prediction models are based on traditional machine learning algorithms and do not fully utilize protein sequence information. To solve this problem, a deep learning model based on self-attention and multiple-channel feature fusion was proposed to predict thermophilic proteins, called DeepTP. First, a large new dataset consisting of 20,842 proteins was constructed. Second, a convolutional neural network and bidirectional long short-term memory network were used to extract the hidden features in protein sequences. Different weights were then assigned to features through self-attention, and finally, biological features were integrated to build a prediction model. In a performance comparison with existing methods, DeepTP had better performance and scalability in an independent balanced test set and validation set, with AUC values of 0.944 and 0.801, respectively. In the unbalanced test set, DeepTP had an average precision (AP) of 0.536. The tool is freely available.

BayeStab: Predicting effects of mutations on protein stability with uncertainty quantification

Predicting protein thermostability change upon mutation is crucial for understanding diseases and designing therapeutics. However, accurately estimating Gibbs free energy change of the protein remained a challenge. Some methods struggle to generalize on examples with no homology and produce uncalibrated predictions. Here we leverage advances in graph neural networks for protein feature extraction to tackle this structure-property prediction task. Our method, BayeStab, is then tested on four test datasets, including S669, S611, S350, and Myoglobin, showing high generalization and symmetry performance. Meanwhile, we apply concrete dropout enabled Bayesian neural networks to infer plausible models and estimate uncertainty. By decomposing the uncertainty into parts induced by data noise and model, we demonstrate that the probabilistic method allows insights into the inherent noise of the training datasets, which is closely relevant to the upper bound of the task. Finally, the BayeStab web server is created and can be found at: http://www.bayestab.com. The code for this work is available at: https://github.com/HongzhouTang/BayeStab.

ProTstab2 for Prediction of Protein Thermal Stabilities

The stability of proteins is an essential property that has several biological implications. Knowledge about protein stability is important in many ways, ranging from protein purification and structure determination to stability in cells and biotechnological applications. Experimental determination of thermal stabilities has been tedious and available data have been limited. The introduction of limited proteolysis and mass spectrometry approaches has facilitated more extensive cellular protein stability data production. We collected melting temperature information for 34,913 proteins and developed a machine learning predictor, ProTstab2, by utilizing a gradient boosting algorithm after testing seven algorithms. The method performance was assessed on a blind test data set and showed a Pearson correlation coefficient of 0.753 and root mean square error of 7.005. Comparison to previous methods indicated that ProTstab2 had superior performance. The method is fast, so it was applied to predict and compare the stabilities of all proteins in human, mouse, and zebrafish proteomes for which experimental data were not determined. The tool is freely available.

Lymphocytes Utilize Somatic Mutations, Epigenetic Silencing, and the Proteasome to Escape Truncated WASP Expression

Wiskott-Aldrich Syndrome Protein (WASP) deficiency causes Wiskott-Aldrich Syndrome (WAS), a sex-linked disorder characterized by combined immunodeficiency, microthrombocytopenia, and eczema. Like WASP-deficient humans, WASP-deficient mice produce normal numbers of functionally defective T cells. Here, we report a WAS patient with a novel germline frameshifting WAS mutation encoding a truncated form of WASP lacking the C-terminal cofilin homology (C) and the acidic region (A) domains (WASPΔCA). Although stably overexpressed in embryonic kidney cell lines, WASPΔCA was undetectable in circulating patient leukocytes. Deep sequencing, transcript profiling, and protein degradation analyses demonstrated patient lymphocytes employ an array of genetic, epigenetic, and proteasomal strategies to avoid expressing WASPΔCA.

A Bibliometric Analysis and Review of Pullulan-Degrading Enzymes—Past and Current Trends

Starch and pullulan degrading enzymes are essential industrial biocatalysts. Pullulan-degrading enzymes are grouped into pullulanases (types I and type II) and pullulan hydrolase (types I, II and III). Generally, these enzymes hydrolyse the α-1,6 glucosidic bonds (and α-1,4 for certain enzyme groups) of substrates and form reducing sugars such as glucose, maltose, maltotriose, panose or isopanose. This review covers two main aspects: (i) bibliometric analysis of publications and patents related to pullulan-degrading enzymes and (ii) biological aspects of free and immobilised pullulan-degrading enzymes and protein engineering. The collective data suggest that most publications involved researchers within the same institution or country in the past and current practice. Multi-national interaction shall be improved, especially in tapping the enzymes from unculturable prokaryotes. While the understanding of pullulanases may reach a certain extend of saturation, the discovery of pullulan hydrolases is still limited. In this report, we suggest readers consider using the next-generation sequencing technique to fill the gaps of finding more new sequences encoding pullulan-degrading enzymes to expand the knowledge body of this topic.

A combined pH and temperature precipitation step facilitates the purification of tobacco-derived recombinant proteins that are sensitive to extremes of either parameter

Incubation at pH 4.0 or blanching at ∼65°C facilitates the purification of biopharmaceutical proteins from plants by precipitating most of the host cell proteins (HCPs) before chromatography. However, both methods are compatible only with pH or thermostable target proteins whereas many target proteins may irreversibly denature, e.g., at pH < 4.0. Here, we developed a combined pH/temperature treatment for clarified tobacco extracts and intact leaves. The latter were subjected to a blanching procedure, i.e., the submersion into a hot buffer. Using a design of experiments approach we identified conditions that remove ∼70% of HCPs at ∼55°C, using the thermosensitive antibody 2G12 and the pH-sensitive DsRed as model proteins. We found that pH and temperature exerted a combined effect during the precipitation of HCPs in the pH range 5.0-7.0 at 35°C-60°C. For clarified extracts, the temperature required to achieve a DsRed purity threshold of 20% total soluble protein (TSP) increased from 54°C to 63°C when the pH was increased from 6.4 to 7.3. The pH-stable antibody 2G12 was less responsive to the combined treatment, but the purity of 1% TSP was achieved at 35°C instead of 44°C when the pH was reduced from 6.3 to 5.8. When blanching intact leaves, product losses were not exacerbated at pH 4.0. Indeed, the highest DsRed purity (58% TSP) was achieved at this pH, combined with a temperature of 60°C and an incubation time of 30 min. In contrast, the highest 2G12 purity (0.7% TSP) was achieved at pH 5.1 and 40°C with an incubation time of 20 min. Our data suggest that a combined pH/temperature regime can avoid extreme values of either parameter; therefore, broadening the applicability of these simple purification techniques to other recombinant proteins.

PLXNA2 and LRRC40 as candidate genes in autism spectrum disorder

Autism spectrum disorder (ASD) is a neurodevelopmental disability with high heritability yet the genetic etiology remains elusive. Therefore, it is necessary to elucidate new genotype-phenotype relationships for ASD to improve both the etiological knowledge and diagnosis. In this work, a copy-number variant and whole-exome sequencing analysis were performed in an ASD patient with a complex neurobehavioral phenotype with epilepsy and attention deficit hyperactivity disorder. We identified rare recessive single nucleotide variants in the two genes, PLXNA2 encoding Plexin A2 that participates in neurodevelopment, and LRRC40, which encodes Leucine-rich repeat containing protein 40, a protein of unknown function. PLXNA2 showed the heterozygous missense variants c.614G>A (p.Arg205Gln) and c.4904G>A (p.Arg1635Gln) while LRRC40 presented the homozygous missense variant c.1461G>T (p.Leu487Phe). In silico analysis predicted that these variants could be pathogenic. We studied PLXNA2 and LRRC40 mRNA and proteins in fibroblasts from the patient and controls. We observed a significant PlxnA2 subcellular delocalization and very low levels of LRRC40 in the patient. Moreover, we found a novel interaction between PlxnA2 and LRRC40 suggesting that participate in a common neural pathway. This interaction was significant decreased in the patient's fibroblasts. In conclusion, our results identified PLXNA2 and LRRC40 genes as candidates in ASD providing novel clues for the pathogenesis. Further attention to these genes is warranted in genetic studies of patients with neurodevelopmental disorders, particularly ASD. LAY SUMMARY: Genomics is improving the knowledge and diagnosis of patients with autism spectrum disorder (ASD) yet the genetic etiology remains elusive. Here, using genomic analysis together with experimental functional studies, we identified in an ASD complex patient the PLXNA2 and LRRC40 recessive genes as ASD candidates. Furthermore, we found that the proteins of these genes interact in a common neural network. Therefore, more attention to these genes is warranted in genetic studies of patients with neurodevelopmental disorders, particularly ASD.

Limitations and challenges in protein stability prediction upon genome variations: towards future applications in precision medicine

Protein stability predictions are becoming essential in medicine to develop novel immunotherapeutic agents and for drug discovery. Despite the large number of computational approaches for predicting the protein stability upon mutation, there are still critical unsolved problems: 1) the limited number of thermodynamic measurements for proteins provided by current databases; 2) the large intrinsic variability of ΔΔG values due to different experimental conditions; 3) biases in the development of predictive methods caused by ignoring the anti-symmetry of ΔΔG values between mutant and native protein forms; 4) over-optimistic prediction performance, due to sequence similarity between proteins used in training and test datasets. Here, we review these issues, highlighting new challenges required to improve current tools and to achieve more reliable predictions. In addition, we provide a perspective of how these methods will be beneficial for designing novel precision medicine approaches for several genetic disorders caused by mutations, such as cancer and neurodegenerative diseases.