ProtTrans: Toward Understanding the Language of Life Through Self-Supervised Learning

TransDTI: Transformer-Based Language Models for Estimating DTIs and Building a Drug Recommendation Workflow

The identification of novel drug-target interactions is a labor-intensive and low-throughput process. In silico alternatives have proved to be of immense importance in assisting the drug discovery process. Here, we present TransDTI, a multiclass classification and regression workflow employing transformer-based language models to segregate interactions between drug-target pairs as active, inactive, and intermediate. The models were trained with large-scale drug-target interaction (DTI) data sets, which reported an improvement in performance in terms of the area under receiver operating characteristic (auROC), the area under precision recall (auPR), Matthew's correlation coefficient (MCC), and R2 over baseline methods. The results showed that models based on transformer-based language models effectively predict novel drug-target interactions from sequence data. The proposed models significantly outperformed existing methods like DeepConvDTI, DeepDTA, and DeepDTI on a test data set. Further, the validity of novel interactions predicted by TransDTI was found to be backed by molecular docking and simulation analysis, where the model prediction had similar or better interaction potential for MAP2k and transforming growth factor-β (TGFβ) and their known inhibitors. Proposed approaches can have a significant impact on the development of personalized therapy and clinical decision making.

Ultrafast end-to-end protein structure prediction enables high-throughput exploration of uncharacterized proteins

Deep learning-based prediction of protein structure usually begins by constructing a multiple sequence alignment (MSA) containing homologs of the target protein. The most successful approaches combine large feature sets derived from MSAs, and considerable computational effort is spent deriving these input features. We present a method that greatly reduces the amount of preprocessing required for a target MSA, while producing main chain coordinates as a direct output of a deep neural network. The network makes use of just three recurrent networks and a stack of residual convolutional layers, making the predictor very fast to run, and easy to install and use. Our approach constructs a directly learned representation of the sequences in an MSA, starting from a one-hot encoding of the sequences. When supplemented with an approximate precision matrix, the learned representation can be used to produce structural models of comparable or greater accuracy as compared to our original DMPfold method, while requiring less than a second to produce a typical model. This level of accuracy and speed allows very large-scale three-dimensional modeling of proteins on minimal hardware, and we demonstrate this by producing models for over 1.3 million uncharacterized regions of proteins extracted from the BFD sequence clusters. After constructing an initial set of approximate models, we select a confident subset of over 30,000 models for further refinement and analysis, revealing putative novel protein folds. We also provide updated models for over 5,000 Pfam families studied in the original DMPfold paper.

Learning protein fitness models from evolutionary and assay-labeled data

Machine learning-based models of protein fitness typically learn from either unlabeled, evolutionarily related sequences or variant sequences with experimentally measured labels. For regimes where only limited experimental data are available, recent work has suggested methods for combining both sources of information. Toward that goal, we propose a simple combination approach that is competitive with, and on average outperforms more sophisticated methods. Our approach uses ridge regression on site-specific amino acid features combined with one probability density feature from modeling the evolutionary data. Within this approach, we find that a variational autoencoder-based probability density model showed the best overall performance, although any evolutionary density model can be used. Moreover, our analysis highlights the importance of systematic evaluations and sufficient baselines.

A guide to machine learning for biologists

The expanding scale and inherent complexity of biological data have encouraged a growing use of machine learning in biology to build informative and predictive models of the underlying biological processes. All machine learning techniques fit models to data; however, the specific methods are quite varied and can at first glance seem bewildering. In this Review, we aim to provide readers with a gentle introduction to a few key machine learning techniques, including the most recently developed and widely used techniques involving deep neural networks. We describe how different techniques may be suited to specific types of biological data, and also discuss some best practices and points to consider when one is embarking on experiments involving machine learning. Some emerging directions in machine learning methodology are also discussed.

Protein inter-residue contact and distance prediction by coupling complementary coevolution features with deep residual networks in CASP14

This article reports and analyzes the results of protein contact and distance prediction by our methods in the 14th Critical Assessment of techniques for protein Structure Prediction (CASP14). A new deep learning-based contact/distance predictor was employed based on the ensemble of two complementary coevolution features coupling with deep residual networks. We also improved our multiple sequence alignment (MSA) generation protocol with wholesale meta-genome sequence databases. On 22 CASP14 free modeling (FM) targets, the proposed model achieved a top-L/5 long-range precision of 63.8% and a mean distance bin error of 1.494. Based on the predicted distance potentials, 11 out of 22 FM targets and all of the 14 FM/template-based modeling (TBM) targets have correctly predicted folds (TM-score >0.5), suggesting that our approach can provide reliable distance potentials for ab initio protein folding.

The generative capacity of probabilistic protein sequence models

Potts models and variational autoencoders (VAEs) have recently gained popularity as generative protein sequence models (GPSMs) to explore fitness landscapes and predict mutation effects. Despite encouraging results, current model evaluation metrics leave unclear whether GPSMs faithfully reproduce the complex multi-residue mutational patterns observed in natural sequences due to epistasis. Here, we develop a set of sequence statistics to assess the "generative capacity" of three current GPSMs: the pairwise Potts Hamiltonian, the VAE, and the site-independent model. We show that the Potts model's generative capacity is largest, as the higher-order mutational statistics generated by the model agree with those observed for natural sequences, while the VAE's lies between the Potts and site-independent models. Importantly, our work provides a new framework for evaluating and interpreting GPSM accuracy which emphasizes the role of higher-order covariation and epistasis, with broader implications for probabilistic sequence models in general.

Computational methods for protein localization prediction

The accurate annotation of protein localization is crucial in understanding protein function in tandem with a broad range of applications such as pathological analysis and drug design. Since most proteins do not have experimentally-determined localization information, the computational prediction of protein localization has been an active research area for more than two decades. In particular, recent machine-learning advancements have fueled the development of new methods in protein localization prediction. In this review paper, we first categorize the main features and algorithms used for protein localization prediction. Then, we summarize a list of protein localization prediction tools in terms of their coverage, characteristics, and accessibility to help users find suitable tools based on their needs. Next, we evaluate some of these tools on a benchmark dataset. Finally, we provide an outlook on the future exploration of protein localization methods.

iCDI-W2vCom: Identifying the Ion Channel-Drug Interaction in Cellular Networking Based on word2vec and node2vec

Ion channels are the second largest drug target family. Ion channel dysfunction may lead to a number of diseases such as Alzheimer's disease, epilepsy, cephalagra, and type II diabetes. In the research work for predicting ion channel-drug, computational approaches are effective and efficient compared with the costly, labor-intensive, and time-consuming experimental methods. Most of the existing methods can only be used to deal with the ion channels of knowing 3D structures; however, the 3D structures of most ion channels are still unknown. Many predictors based on protein sequence were developed to address the challenge, while most of their results need to be improved, or predicting web servers are missing. In this paper, a sequence-based classifier, called "iCDI-W2vCom," was developed to identify the interactions between ion channels and drugs. In the predictor, the drug compound was formulated by SMILES-word2vec, FP2-word2vec, SMILES-node2vec, and ECFPs via a 1184D vector, ion channel was represented by the word2vec via a 64D vector, and the prediction engine was operated by the LightGBM classifier. The accuracy and AUC achieved by iCDI-W2vCom via the fivefold cross validation were 91.95% and 0.9703, which outperformed other existing predictors in this area. A user-friendly web server for iCDI-W2vCom was established at http://www.jci-bioinfo.cn/icdiw2v. The proposed method may also be a potential method for predicting target-drug interaction.

PredictProtein - Predicting Protein Structure and Function for 29 Years

Since 1992 PredictProtein (https://predictprotein.org) is a one-stop online resource for protein sequence analysis with its main site hosted at the Luxembourg Centre for Systems Biomedicine (LCSB) and queried monthly by over 3,000 users in 2020. PredictProtein was the first Internet server for protein predictions. It pioneered combining evolutionary information and machine learning. Given a protein sequence as input, the server outputs multiple sequence alignments, predictions of protein structure in 1D and 2D (secondary structure, solvent accessibility, transmembrane segments, disordered regions, protein flexibility, and disulfide bridges) and predictions of protein function (functional effects of sequence variation or point mutations, Gene Ontology (GO) terms, subcellular localization, and protein-, RNA-, and DNA binding). PredictProtein's infrastructure has moved to the LCSB increasing throughput; the use of MMseqs2 sequence search reduced runtime five-fold (apparently without lowering performance of prediction methods); user interface elements improved usability, and new prediction methods were added. PredictProtein recently included predictions from deep learning embeddings (GO and secondary structure) and a method for the prediction of proteins and residues binding DNA, RNA, or other proteins. PredictProtein.org aspires to provide reliable predictions to computational and experimental biologists alike. All scripts and methods are freely available for offline execution in high-throughput settings.

Evotuning protocols for Transformer-based variant effect prediction on multi-domain proteins

Accurate variant effect prediction has broad impacts on protein engineering. Recent machine learning approaches toward this end are based on representation learning, by which feature vectors are learned and generated from unlabeled sequences. However, it is unclear how to effectively learn evolutionary properties of an engineering target protein from homologous sequences, taking into account the protein's sequence-level structure called domain architecture (DA). Additionally, no optimal protocols are established for incorporating such properties into Transformer, the neural network well-known to perform the best in natural language processing research. This article proposes DA-aware evolutionary fine-tuning, or 'evotuning', protocols for Transformer-based variant effect prediction, considering various combinations of homology search, fine-tuning and sequence vectorization strategies. We exhaustively evaluated our protocols on diverse proteins with different functions and DAs. The results indicated that our protocols achieved significantly better performances than previous DA-unaware ones. The visualizations of attention maps suggested that the structural information was incorporated by evotuning without direct supervision, possibly leading to better prediction accuracy.

Learned Embeddings from Deep Learning to Visualize and Predict Protein Sets

Models from machine learning (ML) or artificial intelligence (AI) increasingly assist in guiding experimental design and decision making in molecular biology and medicine. Recently, Language Models (LMs) have been adapted from Natural Language Processing (NLP) to encode the implicit language written in protein sequences. Protein LMs show enormous potential in generating descriptive representations (embeddings) for proteins from just their sequences, in a fraction of the time with respect to previous approaches, yet with comparable or improved predictive ability. Researchers have trained a variety of protein LMs that are likely to illuminate different angles of the protein language. By leveraging the bio_embeddings pipeline and modules, simple and reproducible workflows can be laid out to generate protein embeddings and rich visualizations. Embeddings can then be leveraged as input features through machine learning libraries to develop methods predicting particular aspects of protein function and structure. Beyond the workflows included here, embeddings have been leveraged as proxies to traditional homology-based inference and even to align similar protein sequences. A wealth of possibilities remain for researchers to harness through the tools provided in the following protocols. © 2021 The Authors. Current Protocols published by Wiley Periodicals LLC. The following protocols are included in this manuscript: Basic Protocol 1: Generic use of the bio_embeddings pipeline to plot protein sequences and annotations Basic Protocol 2: Generate embeddings from protein sequences using the bio_embeddings pipeline Basic Protocol 3: Overlay sequence annotations onto a protein space visualization Basic Protocol 4: Train a machine learning classifier on protein embeddings Alternate Protocol 1: Generate 3D instead of 2D visualizations Alternate Protocol 2: Visualize protein solubility instead of protein subcellular localization Support Protocol: Join embedding generation and sequence space visualization in a pipeline.

Pretraining model for biological sequence data

With the development of high-throughput sequencing technology, biological sequence data reflecting life information becomes increasingly accessible. Particularly on the background of the COVID-19 pandemic, biological sequence data play an important role in detecting diseases, analyzing the mechanism and discovering specific drugs. In recent years, pretraining models that have emerged in natural language processing have attracted widespread attention in many research fields not only to decrease training cost but also to improve performance on downstream tasks. Pretraining models are used for embedding biological sequence and extracting feature from large biological sequence corpus to comprehensively understand the biological sequence data. In this survey, we provide a broad review on pretraining models for biological sequence data. Moreover, we first introduce biological sequences and corresponding datasets, including brief description and accessible link. Subsequently, we systematically summarize popular pretraining models for biological sequences based on four categories: CNN, word2vec, LSTM and Transformer. Then, we present some applications with proposed pretraining models on downstream tasks to explain the role of pretraining models. Next, we provide a novel pretraining scheme for protein sequences and a multitask benchmark for protein pretraining models. Finally, we discuss the challenges and future directions in pretraining models for biological sequences.

Cellular Chaperone Function of Intrinsically Disordered Dehydrin ERD14

Disordered plant chaperones play key roles in helping plants survive in harsh conditions, and they are indispensable for seeds to remain viable. Aside from well-known and thoroughly characterized globular chaperone proteins, there are a number of intrinsically disordered proteins (IDPs) that can also serve as highly effective protecting agents in the cells. One of the largest groups of disordered chaperones is the group of dehydrins, proteins that are expressed at high levels under different abiotic stress conditions, such as drought, high temperature, or osmotic stress. Dehydrins are characterized by the presence of different conserved sequence motifs that also serve as the basis for their categorization. Despite their accepted importance, the exact role and relevance of the conserved regions have not yet been formally addressed. Here, we explored the involvement of each conserved segment in the protective function of the intrinsically disordered stress protein (IDSP) A. thaliana's Early Response to Dehydration (ERD14). We show that segments that are directly involved in partner binding, and others that are not, are equally necessary for proper function and that cellular protection emerges from the balanced interplay of different regions of ERD14.

Representation learning applications in biological sequence analysis

Although remarkable advances have been reported in high-throughput sequencing, the ability to aptly analyze a substantial amount of rapidly generated biological (DNA/RNA/protein) sequencing data remains a critical hurdle. To tackle this issue, the application of natural language processing (NLP) to biological sequence analysis has received increased attention. In this method, biological sequences are regarded as sentences while the single nucleic acids/amino acids or k-mers in these sequences represent the words. Embedding is an essential step in NLP, which performs the conversion of these words into vectors. Specifically, representation learning is an approach used for this transformation process, which can be applied to biological sequences. Vectorized biological sequences can then be applied for function and structure estimation, or as input for other probabilistic models. Considering the importance and growing trend for the application of representation learning to biological research, in the present study, we have reviewed the existing knowledge in representation learning for biological sequence analysis.

Protein transfer learning improves identification of heat shock protein families

Heat shock proteins (HSPs) play a pivotal role as molecular chaperones against unfavorable conditions. Although HSPs are of great importance, their computational identification remains a significant challenge. Previous studies have two major limitations. First, they relied heavily on amino acid composition features, which inevitably limited their prediction performance. Second, their prediction performance was overestimated because of the independent two-stage evaluations and train-test data redundancy. To overcome these limitations, we introduce two novel deep learning algorithms: (1) time-efficient DeepHSP and (2) high-performance DeeperHSP. We propose a convolutional neural network (CNN)-based DeepHSP that classifies both non-HSPs and six HSP families simultaneously. It outperforms state-of-the-art algorithms, despite taking 14–15 times less time for both training and inference. We further improve the performance of DeepHSP by taking advantage of protein transfer learning. While DeepHSP is trained on raw protein sequences, DeeperHSP is trained on top of pre-trained protein representations. Therefore, DeeperHSP remarkably outperforms state-of-the-art algorithms increasing F1 scores in both cross-validation and independent test experiments by 20% and 10%, respectively. We envision that the proposed algorithms can provide a proteome-wide prediction of HSPs and help in various downstream analyses for pathology and clinical research.

MSA-Regularized Protein Sequence Transformer toward Predicting Genome-Wide Chemical-Protein Interactions: Application to GPCRome Deorphanization

Small molecules play a critical role in modulating biological systems. Knowledge of chemical-protein interactions helps address fundamental and practical questions in biology and medicine. However, with the rapid emergence of newly sequenced genes, the endogenous or surrogate ligands of a vast number of proteins remain unknown. Homology modeling and machine learning are two major methods for assigning new ligands to a protein but mostly fail when sequence homology between an unannotated protein and those with known functions or structures is low. In this study, we develop a new deep learning framework to predict chemical binding to evolutionary divergent unannotated proteins, whose ligand cannot be reliably predicted by existing methods. By incorporating evolutionary information into self-supervised learning of unlabeled protein sequences, we develop a novel method, distilled sequence alignment embedding (DISAE), for the protein sequence representation. DISAE can utilize all protein sequences and their multiple sequence alignment (MSA) to capture functional relationships between proteins without the knowledge of their structure and function. Followed by the DISAE pretraining, we devise a module-based fine-tuning strategy for the supervised learning of chemical-protein interactions. In the benchmark studies, DISAE significantly improves the generalizability of machine learning models and outperforms the state-of-the-art methods by a large margin. Comprehensive ablation studies suggest that the use of MSA, sequence distillation, and triplet pretraining critically contributes to the success of DISAE. The interpretability analysis of DISAE suggests that it learns biologically meaningful information. We further use DISAE to assign ligands to human orphan G-protein coupled receptors (GPCRs) and to cluster the human GPCRome by integrating their phylogenetic and ligand relationships. The promising results of DISAE open an avenue for exploring the chemical landscape of entire sequenced genomes.

The language of proteins: NLP, machine learning & protein sequences

Natural language processing (NLP) is a field of computer science concerned with automated text and language analysis. In recent years, following a series of breakthroughs in deep and machine learning, NLP methods have shown overwhelming progress. Here, we review the success, promise and pitfalls of applying NLP algorithms to the study of proteins. Proteins, which can be represented as strings of amino-acid letters, are a natural fit to many NLP methods. We explore the conceptual similarities and differences between proteins and language, and review a range of protein-related tasks amenable to machine learning. We present methods for encoding the information of proteins as text and analyzing it with NLP methods, reviewing classic concepts such as bag-of-words, k-mers/n-grams and text search, as well as modern techniques such as word embedding, contextualized embedding, deep learning and neural language models. In particular, we focus on recent innovations such as masked language modeling, self-supervised learning and attention-based models. Finally, we discuss trends and challenges in the intersection of NLP and protein research.

Improved protein structure refinement guided by deep learning based accuracy estimation

We develop a deep learning framework (DeepAccNet) that estimates per-residue accuracy and residue-residue distance signed error in protein models and uses these predictions to guide Rosetta protein structure refinement. The network uses 3D convolutions to evaluate local atomic environments followed by 2D convolutions to provide their global contexts and outperforms other methods that similarly predict the accuracy of protein structure models. Overall accuracy predictions for X-ray and cryoEM structures in the PDB correlate with their resolution, and the network should be broadly useful for assessing the accuracy of both predicted structure models and experimentally determined structures and identifying specific regions likely to be in error. Incorporation of the accuracy predictions at multiple stages in the Rosetta refinement protocol considerably increased the accuracy of the resulting protein structure models, illustrating how deep learning can improve search for global energy minima of biomolecules.

Generating functional protein variants with variational autoencoders

The vast expansion of protein sequence databases provides an opportunity for new protein design approaches which seek to learn the sequence-function relationship directly from natural sequence variation. Deep generative models trained on protein sequence data have been shown to learn biologically meaningful representations helpful for a variety of downstream tasks, but their potential for direct use in the design of novel proteins remains largely unexplored. Here we show that variational autoencoders trained on a dataset of almost 70000 luciferase-like oxidoreductases can be used to generate novel, functional variants of the luxA bacterial luciferase. We propose separate VAE models to work with aligned sequence input (MSA VAE) and raw sequence input (AR-VAE), and offer evidence that while both are able to reproduce patterns of amino acid usage characteristic of the family, the MSA VAE is better able to capture long-distance dependencies reflecting the influence of 3D structure. To confirm the practical utility of the models, we used them to generate variants of luxA whose luminescence activity was validated experimentally. We further showed that conditional variants of both models could be used to increase the solubility of luxA without disrupting function. Altogether 6/12 of the variants generated using the unconditional AR-VAE and 9/11 generated using the unconditional MSA VAE retained measurable luminescence, together with all 23 of the less distant variants generated by conditional versions of the models; the most distant functional variant contained 35 differences relative to the nearest training set sequence. These results demonstrate the feasibility of using deep generative models to explore the space of possible protein sequences and generate useful variants, providing a method complementary to rational design and directed evolution approaches.

Embeddings from deep learning transfer GO annotations beyond homology

Knowing protein function is crucial to advance molecular and medical biology, yet experimental function annotations through the Gene Ontology (GO) exist for fewer than 0.5% of all known proteins. Computational methods bridge this sequence-annotation gap typically through homology-based annotation transfer by identifying sequence-similar proteins with known function or through prediction methods using evolutionary information. Here, we propose predicting GO terms through annotation transfer based on proximity of proteins in the SeqVec embedding rather than in sequence space. These embeddings originate from deep learned language models (LMs) for protein sequences (SeqVec) transferring the knowledge gained from predicting the next amino acid in 33 million protein sequences. Replicating the conditions of CAFA3, our method reaches an Fmax of 37 ± 2%, 50 ± 3%, and 57 ± 2% for BPO, MFO, and CCO, respectively. Numerically, this appears close to the top ten CAFA3 methods. When restricting the annotation transfer to proteins with < 20% pairwise sequence identity to the query, performance drops (Fmax BPO 33 ± 2%, MFO 43 ± 3%, CCO 53 ± 2%); this still outperforms naïve sequence-based transfer. Preliminary results from CAFA4 appear to confirm these findings. Overall, this new concept is likely to change the annotation of proteins, in particular for proteins from smaller families or proteins with intrinsically disordered regions.

A Review of Protein Structure Prediction using Deep Learning

Proteins are macromolecules composed of 20 types of amino acids in a specific order. Understanding how proteins fold is vital because its 3-dimensional structure determines the function of a protein. Prediction of protein structure based on amino acid strands and evolutionary information becomes the basis for other studies such as predicting the function, property or behaviour of a protein and modifying or designing new proteins to perform certain desired functions. Machine learning advances, particularly deep learning, are igniting a paradigm shift in scientific study. In this review, we summarize recent work in applying deep learning techniques to tackle problems in protein structural prediction. We discuss various deep learning approaches used to predict protein structure and future achievements and challenges. This review is expected to help provide perspectives on problems in biochemistry that can take advantage of the deep learning approach. Some of the unanswered challenges with current computational approaches are predicting the location and precision orientation of protein side chains, predicting protein interactions with DNA, RNA and other small molecules and predicting the structure of protein complexes.