Search

The protein inverse folding problem, designing amino acid sequences that fold into desired protein structures, is a critical challenge in biological sciences. Despite numerous data-driven and knowledge-driven methods, there remains a need for a user-friendly toolkit that effectively integrates these approaches for in-silico protein design. In this paper, we present DIProT, an interactive protein design toolkit. DIProT leverages a non-autoregressive deep generative model to solve the inverse folding problem, combined with a protein structure prediction model. This integration allows users to incorporate prior knowledge into the design process, evaluate designs in silico, and form a virtual design loop with human feedback. Our inverse folding model demonstrates competitive performance in terms of effectiveness and efficiency on TS50 and CATH4.2 datasets, with promising sequence recovery and inference time. Case studies further illustrate how DIProT can facilitate user-guided protein design. [ABSTRACT FROM AUTHOR]

Featured Application: Protein and peptide identification based on tandem mass spectrometry is a pillar technology in proteomics research. In recent years, increasing numbers of researchers have utilized deep learning to tackle challenges in proteomics. For example, catalyzed by deep learning, AlphaFold has achieved unparalleled levels of accuracy in protein-structure prediction. Prior to studying the structure and function of proteins in cells or tissues, it is essential to determine the sequences of amino acids in peptides or proteins. De novo peptide sequencing can be used to directly infer the peptide sequence from a tandem mass spectrum without the requirement for a reference sequence database, making it particularly suitable for the determination of protein sequences of unknown species, monoclonal antibodies, and cancer neoantigens. The de novo peptide-sequencing method can be used to directly infer the peptide sequence from a tandem mass spectrum. It has the advantage of not relying on protein databases and plays a key role in the determination of the protein sequences of unknown species, monoclonal antibodies, and cancer neoantigens. In this paper, we propose a method based on graph convolutional neural networks and convolutional neural networks, Denovo-GCN, for de novo peptide sequencing. We constructed an undirected graph based on the mass difference between the spectral peaks in a tandem mass spectrum. The features of the nodes on the spectrum graph, which represent the spectral peaks, were the matching information of the peptide sequence and the mass spectrum. Next, the Denovo-GCN used CNN to extract the features of the nodes. The correlation between the nodes was represented by an adjacency matrix, which aggregated the features of neighboring nodes. Denovo-GCN provides a complete end-to-end training and prediction framework to sequence patterns of peptides. Our experiments on various data sets from different species show that Denovo-GCN outperforms DeepNovo with a relative improvement of 13.7–25.5% in terms of the peptide-level recall. [ABSTRACT FROM AUTHOR]

The X-ray diffraction (XRD) technique based on crystallography is the main experimental method to analyze the three-dimensional structure of proteins. The production process of protein crystals on which the XRD technique relies has undergone multiple experimental steps, which requires a lot of manpower and material resources. In addition, studies have shown that not all proteins can form crystals under experimental conditions, and the success rate of the final crystallization of proteins is only <10%. Although some protein crystallization predictors have been developed, not many tools capable of predicting multi-stage protein crystallization propensity are available and the accuracy of these tools is not satisfactory. In this paper, we propose a novel deep learning framework, named SADeepcry, for predicting protein crystallization propensity. The framework can be used to estimate the three steps (protein material production, purification and crystallization) in protein crystallization experiments and the success rate of the final protein crystallization. SADeepcry uses the optimized self-attention and auto-encoder modules to extract sequence, structure and physicochemical features from the proteins. Compared with other state-of-the-art protein crystallization propensity prediction models, SADeepcry can obtain more complex global spatial long-distance dependence of protein sequence information. Our computational results show that SADeepcry has increased Matthews correlation coefficient and area under the curve, by 100.3% and 13.4%, respectively, over the DCFCrystal method on the benchmark dataset. The codes of SADeepcry are available at https://github.com/zhc940702/SADeepcry. [ABSTRACT FROM AUTHOR]

Understanding gene functions and their associated abnormal phenotypes is crucial in the prevention, diagnosis and treatment against diseases. The Human Phenotype Ontology (HPO) is a standardized vocabulary for describing the phenotype abnormalities associated with human diseases. However, the current HPO annotations are far from completion, and only a small fraction of human protein-coding genes has HPO annotations. Thus, it is necessary to predict protein-phenotype associations using computational methods. Protein sequences can indicate the structure and function of the proteins, and interacting proteins are more likely to have same function. It is promising to integrate these features for predicting HPO annotations of human protein. We developed GraphPheno, a semi-supervised method based on graph autoencoders, which does not require feature engineering to capture deep features from protein sequences, while also taking into account the topological properties in the protein–protein interaction network to predict the relationships between human genes/proteins and abnormal phenotypes. Cross validation and independent dataset tests show that GraphPheno has satisfactory prediction performance. The algorithm is further confirmed on automatic HPO annotation for no-knowledge proteins under the benchmark of the second Critical Assessment of Functional Annotation, 2013–2014 (CAFA2), where GraphPheno surpasses most existing methods. Further bioinformatics analysis shows that predicted certain phenotype-associated genes using GraphPheno share similar biological properties with known ones. In a case study on the phenotype of abnormality of mitochondrial respiratory chain, top prioritized genes are validated by recent papers. We believe that GraphPheno will help to reveal more associations between genes and phenotypes, and contribute to the discovery of drug targets. [ABSTRACT FROM AUTHOR]

The solubility of proteins stands as a pivotal factor in the realm of pharmaceutical research and production. Addressing the imperative to enhance production efficiency and curtail experimental costs, the demand arises for computational models adept at accurately predicting solubility based on provided datasets. Prior investigations have leveraged deep learning models and feature engineering techniques to distill features from raw protein sequences for solubility prediction. However, these methodologies have not thoroughly delved into the interdependencies among features or their respective magnitudes of significance. This study introduces HybridGCN, a pioneering Hybrid Graph Convolutional Network that elevates solubility prediction accuracy through the combination of diverse features, encompassing sophisticated deep-learning features and classical biophysical features. An exploration into the intricate interplay between deep-learning features and biophysical features revealed that specific biophysical attributes, notably evolutionary features, complement features extracted by advanced deep-learning models. Augmenting the model's capability for feature representation, we employed ESM, a substantial protein language model, to derive a zero-shot learning feature capturing comprehensive and pertinent information concerning protein functions and structures. Furthermore, we proposed a novel feature fusion module termed Adaptive Feature Re-weighting (AFR) to integrate multiple features, thereby enabling the fine-tuning of feature importance. Ablation experiments and comparative analyses attest to the efficacy of the HybridGCN approach, culminating in state-of-the-art performances on the public eSOL and S. cerevisiae datasets. [ABSTRACT FROM AUTHOR]

Background: Intrinsically disordered regions (IDRs) are widely distributed in proteins and related to many important biological functions. Accurately identifying IDRs is of great significance for protein structure and function analysis. Because the long disordered regions (LDRs) and short disordered regions (SDRs) share different characteristics, the existing predictors fail to achieve better and more stable performance on datasets with different ratios between LDRs and SDRs. There are two main reasons. First, the existing predictors construct network structures based on their own experiences such as convolutional neural network (CNN) which is used to extract the feature of neighboring residues in protein, and long short-term memory (LSTM) is used to extract the long-distance dependencies feature of protein residues. But these networks cannot capture the hidden feature associated with the length-dependent between residues. Second, many algorithms based on deep learning have been proposed but the complementarity of the existing predictors is not fully explored and used. Results: In this study, the neural architecture search (NAS) algorithm was employed to automatically construct the network structures so as to capture the hidden features in protein sequences. In order to stably predict both the LDRs and SDRs, the model constructed by NAS was combined with length-dependent models for capturing the unique features of SDRs or LDRs and general models for capturing the common features between LDRs and SDRs. A new predictor called IDP-Fusion was proposed. Conclusions: Experimental results showed that IDP-Fusion can achieve more stable performance than the other existing predictors on independent test sets with different ratios between SDRs and LDRs. [ABSTRACT FROM AUTHOR]

Nuclear Magnetic Resonance (NMR) spectroscopy is a major technique in structural biology with over 11,800 protein structures deposited in the Protein Data Bank. NMR can elucidate structures and dynamics of small and medium size proteins in solution, living cells, and solids, but has been limited by the tedious data analysis process. It typically requires weeks or months of manual work of a trained expert to turn NMR measurements into a protein structure. Automation of this process is an open problem, formulated in the field over 30 years ago. We present a solution to this challenge that enables the completely automated analysis of protein NMR data within hours after completing the measurements. Using only NMR spectra and the protein sequence as input, our machine learning-based method, ARTINA, delivers signal positions, resonance assignments, and structures strictly without human intervention. Tested on a 100-protein benchmark comprising 1329 multidimensional NMR spectra, ARTINA demonstrated its ability to solve structures with 1.44 Å median RMSD to the PDB reference and to identify 91.36% correct NMR resonance assignments. ARTINA can be used by non-experts, reducing the effort for a protein assignment or structure determination by NMR essentially to the preparation of the sample and the spectra measurements. The analysis of protein NMR spectra is time-consuming and can occupy a human expert for weeks or months. The researchers in this work present a deep learning-based method that delivers signal positions, chemical shift assignments, and structures of proteins within hours after completion of the NMR measurements. [ABSTRACT FROM AUTHOR]

Proteins are constructed from amino acid sequences. Their structural classifications include primary, secondary, tertiary, and quaternary, with tertiary and quaternary structures influencing protein function. Because a protein's structure is inextricably connected to its biological function, machine learning algorithms that can better anticipate the structures have the potential to lead to new scientific discoveries in human health and improve our capacity to develop new treatments. Protein secondary structure assignment enriches the structural and functional understanding of proteins. It helps in protein structure comparison and classification studies, besides facilitating secondary and tertiary structure prediction systems. Several secondary structure assignment methods have been developed since the 1980s, most of which are based on hydrogen bond analysis and atomic coordinate features. However, the assignment process becomes complex when protein data includes missing atoms. Deep neural networks are often referred to as universal function approximators because they can approximate any function to produce the desired output when properly designed and trained. Optimised deep learning architectures have already proven their ability to increase performance in a wide range of problems. Recently, the ResNet architecture has garnered significant interest due to its applicability in various areas, including image classification and protein contact map prediction. The proposed model, which is based on the ResNet architecture, assigns secondary structures using Cα atom coordinates. The model achieved an accuracy of 94% when evaluated against the benchmark and independent test sets. The findings encourage the development of new deep learning-based methods that are more generalised across various protein learning tasks. Furthermore, it allows computational biologists to delve deeper into integrating these techniques with experimental methods. The model codes are available at: https://github.com/jisnava/ResNet_for_Structure_Assignments/. [ABSTRACT FROM AUTHOR]

The information about the domain architecture of proteins is useful for studying protein structure and function. However, accurate prediction of protein domain boundaries (i.e., sequence regions separating two domains) from sequence remains a significant challenge. In this work, we develop a deep learning method based on multi-head U-Nets (called DistDom) to predict protein domain boundaries utilizing 1D sequence features and predicted 2D inter-residue distance map as input. The 1D features contain the evolutionary and physicochemical information of protein sequences, whereas the 2D distance map includes the structural information of proteins that was rarely used in domain boundary prediction before. The 1D and 2D features are processed by the 1D and 2D U-Nets respectively to generate hidden features. The hidden features are then used by the multi-head attention to predict the probability of each residue of a protein being in a domain boundary, leveraging both local and global information in the features. The residue-level domain boundary predictions can be used to classify proteins as single-domain or multi-domain proteins. It classifies the CASP14 single-domain and multi-domain targets at the accuracy of 75.9%, 13.28% more accurate than the state-of-the-art method. Tested on the CASP14 multi-domain protein targets with expert annotated domain boundaries, the average per-target F1 measure score of the domain boundary prediction by DistDom is 0.263, 29.56% higher than the state-of-the-art method. [ABSTRACT FROM AUTHOR]

The mapping from protein sequence to function is highly complex, making it challenging to predict how sequence changes will affect a protein’s behavior and properties. We present a supervised deep learning framework to learn the sequence–function mapping from deep mutational scanning data and make predictions for new, uncharacterized sequence variants. We test multiple neural network architectures, including a graph convolutional network that incorporates protein structure, to explore how a network’s internal representation affects its ability to learn the sequence–function mapping. Our supervised learning approach displays superior performance over physics-based and unsupervised prediction methods. We find that networks that capture nonlinear interactions and share parameters across sequence positions are important for learning the relationship between sequence and function. Further analysis of the trained models reveals the networks’ ability to learn biologically meaningful information about protein structure and mechanism. Finally, we demonstrate the models’ ability to navigate sequence space and design new proteins beyond the training set. We applied the protein G B1 domain (GB1) models to design a sequence that binds to immunoglobulin G with substantially higher affinity than wild-type GB1. [ABSTRACT FROM AUTHOR]

Although genes carry information, proteins are the main role player in providing all the functionalities of a living organism. Massive amounts of different proteins involve in every function that occurs in a cell. These amino acid sequences can be hierarchically classified into a set of families and subfamilies depending on their evolutionary relatedness and similarities in their structure or function. Protein characterization to identify protein structure and function is done accurately using laboratory experiments. With the rapidly increasing huge amount of novel protein sequences, these experiments have become difficult to carry out since they are expensive, time-consuming, and laborious. Therefore, many computational classification methods are introduced to classify proteins and predict their functional properties. With the progress of the performance of the computational techniques, deep learning plays a key role in many areas. Novel deep learning models such as DeepFam, ProtCNN have been presented to classify proteins into their families recently. However, these deep learning models have been used to carry out the non-hierarchical classification of proteins. In this research, we propose a deep learning neural network model named DeepHiFam with high accuracy to classify proteins hierarchically into different levels simultaneously. The model achieved an accuracy of 98.38% for protein family classification and more than 80% accuracy for the classification of protein subfamilies and sub-subfamilies. Further, DeepHiFam performed well in the non-hierarchical classification of protein families and achieved an accuracy of 98.62% and 96.14% for the popular Pfam dataset and COG dataset respectively. [ABSTRACT FROM AUTHOR]

Background: Optimization of DNA and protein sequences based on Machine Learning models is becoming a powerful tool for molecular design. Activation maximization offers a simple design strategy for differentiable models: one-hot coded sequences are first approximated by a continuous representation, which is then iteratively optimized with respect to the predictor oracle by gradient ascent. While elegant, the current version of the method suffers from vanishing gradients and may cause predictor pathologies leading to poor convergence. Results: Here, we introduce Fast SeqProp, an improved activation maximization method that combines straight-through approximation with normalization across the parameters of the input sequence distribution. Fast SeqProp overcomes bottlenecks in earlier methods arising from input parameters becoming skewed during optimization. Compared to prior methods, Fast SeqProp results in up to 100-fold faster convergence while also finding improved fitness optima for many applications. We demonstrate Fast SeqProp's capabilities by designing DNA and protein sequences for six deep learning predictors, including a protein structure predictor. Conclusions: Fast SeqProp offers a reliable and efficient method for general-purpose sequence optimization through a differentiable fitness predictor. As demonstrated on a variety of deep learning models, the method is widely applicable, and can incorporate various regularization techniques to maintain confidence in the sequence designs. As a design tool, Fast SeqProp may aid in the development of novel molecules, drug therapies and vaccines. [ABSTRACT FROM AUTHOR]

The subcellular locations of proteins are closely related to their functions. In the past few decades, the application of machine learning algorithms to predict protein subcellular locations has been an important topic in proteomics. However, most studies in this field used only amino acid sequences as the data source. Only a few works focused on other protein data types. For example, three-dimensional structures, which contain far more functional protein information than sequences, remain to be explored. In this work, we extracted various handcrafted features to describe the protein structures from physical, chemical, and topological aspects, as well as the learned features obtained by deep neural networks. We then used these features to classify the protein subcellular locations. Our experimental results demonstrated that some of these structural features have a certain effect on the protein location classification, and can help improve the performance of sequence-based location predictors. Our method provides a new view for the analysis of protein spatial distribution, and is anticipated to be used in revealing the relationships between protein structures and functions. [ABSTRACT FROM AUTHOR]

Computational Protein Design (CPD) has produced impressive results for engineering new proteins, resulting in a wide variety of applications. In the past few years, various efforts have aimed at replacing or improving existing design methods using Deep Learning technology to leverage the amount of publicly available protein data. Deep Learning (DL) is a very powerful tool to extract patterns from raw data, provided that data are formatted as mathematical objects and the architecture processing them is well suited to the targeted problem. In the case of protein data, specific representations are needed for both the amino acid sequence and the protein structure in order to capture respectively 1D and 3D information. As no consensus has been reached about the most suitable representations, this review describes the representations used so far, discusses their strengths and weaknesses, and details their associated DL architecture for design and related tasks. [ABSTRACT FROM AUTHOR]

Protein secondary structures are important in many biological processes and applications. Due to advances in sequencing methods, there are many proteins sequenced, but fewer proteins with secondary structures defined by laboratory methods. With the development of computer technology, computational methods have (started to) become the most important methodologies for predicting secondary structures. We evaluated two different approaches to this problem—driven by the recent results obtained by computational methods in this task—(i) template-free classifiers, based on machine learning techniques; and (ii) template-based classifiers, based on searching tools. Both approaches are formed by different sub-classifiers—six for template-free and two for template-based, each with a specific view of the protein. Our results show that these ensembles improve the results of each approach individually. [ABSTRACT FROM AUTHOR]