Review of unsupervised pretraining strategies for molecules representation

Positional embeddings and zero-shot learning using BERT for molecular-property prediction

Recently, advancements in cheminformatics such as representation learning for chemical structures, deep learning (DL) for property prediction, data-driven discovery, and optimization of chemical data handling, have led to increased demands for handling chemical simplified molecular input line entry system (SMILES) data, particularly in text analysis tasks. These advancements have driven the need to optimize components like positional encoding and positional embeddings (PEs) in transformer model to better capture the sequential and contextual information embedded in molecular representations. SMILES data represent complex relationships among atoms or elements, rendering them critical for various learning tasks within the field of cheminformatics. This study addresses the critical challenge of encoding complex relationships among atoms in SMILES strings to explore various PEs within the transformer-based framework to increase the accuracy and generalization of molecular property predictions. The success of transformer-based models, such as the bidirectional encoder representations from transformer (BERT) models, in natural language processing tasks has sparked growing interest from the domain of cheminformatics. However, the performance of these models during pretraining and fine-tuning is significantly influenced by positional information such as PEs, which help in understanding the intricate relationships within sequences. Integrating position information within transformer architectures has emerged as a promising approach. This encoding mechanism provides essential supervision for modeling dependencies among elements situated at different positions within a given sequence. In this study, we first conduct pretraining experiments using various PEs to explore diverse methodologies for incorporating positional information into the BERT model for chemical text analysis using SMILES strings. Next, for each PE, we fine-tune the best-performing BERT (masked language modeling) model on downstream tasks for molecular-property prediction. Here, we use two molecular representations, SMILES and DeepSMILES, to comprehensively assess the potential and limitations of the PEs in zero-shot learning analysis, demonstrating the model's proficiency in predicting properties of unseen molecular representations in the context of newly proposed and existing datasets.Scientific contributionThis study explores the unexplored potential of PEs using BERT model for molecular property prediction. The study involved pretraining and fine-tuning the BERT model on various datasets related to COVID-19, bioassay data, and other molecular and biological properties using SMILES and DeepSMILES representations. The study details the pretraining architecture, fine-tuning datasets, and the performance of the BERT model with different PEs. It also explores zero-shot learning analysis and the model's performance on various classification and regression tasks. In this study, newly proposed datasets from different domains were introduced during fine-tuning in addition to the existing and commonly used datasets. The study highlights the robustness of the BERT model in predicting chemical properties and its potential applications in cheminformatics and bioinformatics.

Transformer-based models for chemical SMILES representation: A comprehensive literature review

Pre-trained chemical language models (CLMs) have attracted increasing attention within the domains of cheminformatics and bioinformatics, inspired by their remarkable success in the natural language processing (NLP) domain such as speech recognition, text analysis, translation, and other objectives associated with language. Furthermore, the vast amount of unlabeled data associated with chemical compounds or molecules has emerged as a crucial research focus, prompting the need for CLMs with reasoning capabilities over such data. Molecular graphs and molecular descriptors are the predominant approaches to representing molecules for property prediction in machine learning (ML). However, Transformer-based LMs have recently emerged as de-facto powerful tools in deep learning (DL), showcasing outstanding performance across various NLP downstream tasks, particularly in text analysis. Within the realm of pre-trained transformer-based LMs such as, BERT (and its variants) and GPT (and its variants) have been extensively explored in the chemical informatics domain. Various learning tasks in cheminformatics such as the text analysis that necessitate handling of chemical SMILES data which contains intricate relations among elements or atoms, have become increasingly prevalent. Whether the objective is predicting molecular reactions or molecular property prediction, there is a growing demand for LMs capable of learning molecular contextual information within SMILES sequences or strings from text inputs (i.e., SMILES). This review provides an overview of the current state-of-the-art of chemical language Transformer-based LMs in chemical informatics for de novo design, and analyses current limitations, challenges, and advantages. Finally, a perspective on future opportunities is provided in this evolving field.

Crystal Composition Transformer: Self-Learning Neural Language Model for Generative and Tinkering Design of Materials

Self-supervised neural language models have recently achieved unprecedented success from natural language processing to learning the languages of biological sequences and organic molecules. These models have demonstrated superior performance in the generation, structure classification, and functional predictions for proteins and molecules with learned representations. However, most of the masking-based pre-trained language models are not designed for generative design, and their black-box nature makes it difficult to interpret their design logic. Here a Blank-filling Language Model for Materials (BLMM) Crystal Transformer is proposed, a neural network-based probabilistic generative model for generative and tinkering design of inorganic materials. The model is built on the blank-filling language model for text generation and has demonstrated unique advantages in learning the "materials grammars" together with high-quality generation, interpretability, and data efficiency. It can generate chemically valid materials compositions with as high as 89.7% charge neutrality and 84.8% balanced electronegativity, which are more than four and eight times higher compared to a pseudo-random sampling baseline. The probabilistic generation process of BLMM allows it to recommend materials tinkering operations based on learned materials chemistry, which makes it useful for materials doping. The model is applied to discover a set of new materials as validated using the Density Functional Theory (DFT) calculations. This work thus brings the unsupervised transformer language models based generative artificial intelligence to inorganic materials. A user-friendly web app for tinkering materials design has been developed and can be accessed freely at www.materialsatlas.org/blmtinker.

DAS-DDI: A dual-view framework with drug association and drug structure for drug-drug interaction prediction

In drug development and clinical application, drug-drug interaction (DDI) prediction is crucial for patient safety and therapeutic efficacy. However, traditional methods for DDI prediction often overlook the structural features of drugs and the complex interrelationships between them, which affect the accuracy and interpretability of the model. In this paper, a novel dual-view DDI prediction framework, DAS-DDI is proposed. Firstly, a drug association network is constructed based on similarity information among drugs, which could provide rich context information for DDI prediction. Subsequently, a novel drug substructure extraction method is proposed, which could update the features of nodes and chemical bonds simultaneously, improving the comprehensiveness of the feature. Furthermore, an attention mechanism is employed to fuse multiple drug embeddings from different views dynamically, enhancing the discriminative ability of the model in handling multi-view data. Comparative experiments on three public datasets demonstrate the superiority of DAS-DDI compared with other state-of-the-art models under two scenarios.

FTMMR: Fusion Transformer for Integrating Multiple Molecular Representations

Molecular property prediction has gained substantial attention due to its potential for various bio-chemical applications. Numerous attempts have been made to enhance the performance by combining multiple molecular representations (1D, 2D, and 3D). However, most prior works only merged a limited number of representations or tried to embed multiple representations through a single network without using representation-specific networks. Furthermore, the heterogeneous characteristics of each representation made the fusion more challenging. Addressing these challenges, we introduce the Fusion Transformer for Multiple Molecular Representations (FTMMR) framework. Our strategy employs three distinct representation-specific networks and integrates information from each network using a fusion transformer architecture to generate fused representations. Additionally, we use self-supervised learning methods to align heterogeneous representations and to effectively utilize the limited chemical data available. In particular, we adopt a combinatorial loss function to leverage the contrastive loss for all three representations. We evaluate the performance of FTMMR using seven benchmark datasets, demonstrating that our framework outperforms existing fusion and self-supervised methods.

MolLM: a unified language model for integrating biomedical text with 2D and 3D molecular representations

MOTIVATION

The current paradigm of deep learning models for the joint representation of molecules and text primarily relies on 1D or 2D molecular formats, neglecting significant 3D structural information that offers valuable physical insight. This narrow focus inhibits the models' versatility and adaptability across a wide range of modalities. Conversely, the limited research focusing on explicit 3D representation tends to overlook textual data within the biomedical domain.

RESULTS

We present a unified pre-trained language model, MolLM, that concurrently captures 2D and 3D molecular information alongside biomedical text. MolLM consists of a text Transformer encoder and a molecular Transformer encoder, designed to encode both 2D and 3D molecular structures. To support MolLM's self-supervised pre-training, we constructed 160K molecule-text pairings. Employing contrastive learning as a supervisory signal for learning, MolLM demonstrates robust molecular representation capabilities across four downstream tasks, including cross-modal molecule and text matching, property prediction, captioning, and text-prompted molecular editing. Through ablation, we demonstrate that the inclusion of explicit 3D representations improves performance in these downstream tasks.

AVAILABILITY AND IMPLEMENTATION

Our code, data, pre-trained model weights, and examples of using our model are all available at https://github.com/gersteinlab/MolLM. In particular, we provide Jupyter Notebooks offering step-by-step guidance on how to use MolLM to extract embeddings for both molecules and text.

Deep learning for low-data drug discovery: Hurdles and opportunities

Deep learning is becoming increasingly relevant in drug discovery, from de novo design to protein structure prediction and synthesis planning. However, it is often challenged by the small data regimes typical of certain drug discovery tasks. In such scenarios, deep learning approaches-which are notoriously 'data-hungry'-might fail to live up to their promise. Developing novel approaches to leverage the power of deep learning in low-data scenarios is sparking great attention, and future developments are expected to propel the field further. This mini-review provides an overview of recent low-data-learning approaches in drug discovery, analyzing their hurdles and advantages. Finally, we venture to provide a forecast of future research directions in low-data learning for drug discovery.

Complementary multi-modality molecular self-supervised learning via non-overlapping masking for property prediction

Self-supervised learning plays an important role in molecular representation learning because labeled molecular data are usually limited in many tasks, such as chemical property prediction and virtual screening. However, most existing molecular pre-training methods focus on one modality of molecular data, and the complementary information of two important modalities, SMILES and graph, is not fully explored. In this study, we propose an effective multi-modality self-supervised learning framework for molecular SMILES and graph. Specifically, SMILES data and graph data are first tokenized so that they can be processed by a unified Transformer-based backbone network, which is trained by a masked reconstruction strategy. In addition, we introduce a specialized non-overlapping masking strategy to encourage fine-grained interaction between these two modalities. Experimental results show that our framework achieves state-of-the-art performance in a series of molecular property prediction tasks, and a detailed ablation study demonstrates efficacy of the multi-modality framework and the masking strategy.

PermuteDDS: a permutable feature fusion network for drug-drug synergy prediction

MOTIVATION

Drug combination therapies have shown promise in clinical cancer treatments. However, it is hard to experimentally identify all drug combinations for synergistic interaction even with high-throughput screening due to the vast space of potential combinations. Although a number of computational methods for drug synergy prediction have proven successful in narrowing down this space, fusing drug pairs and cell line features effectively still lacks study, hindering current algorithms from understanding the complex interaction between drugs and cell lines.

RESULTS

In this paper, we proposed a Permutable feature fusion network for Drug-Drug Synergy prediction, named PermuteDDS. PermuteDDS takes multiple representations of drugs and cell lines as input and employs a permutable fusion mechanism to combine drug and cell line features. In experiments, PermuteDDS exhibits state-of-the-art performance on two benchmark data sets. Additionally, the results on independent test set grouped by different tissues reveal that PermuteDDS has good generalization performance. We believed that PermuteDDS is an effective and valuable tool for identifying synergistic drug combinations. It is publicly available at https://github.com/littlewei-lazy/PermuteDDS .

SCIENTIFIC CONTRIBUTION

First, this paper proposes a permutable feature fusion network for predicting drug synergy termed PermuteDDS, which extract diverse information from multiple drug representations and cell line representations. Second, the permutable fusion mechanism combine the drug and cell line features by integrating information of different channels, enabling the utilization of complex relationships between drugs and cell lines. Third, comparative and ablation experiments provide evidence of the efficacy of PermuteDDS in predicting drug-drug synergy.

A deep learning framework for predicting molecular property based on multi-type features fusion

Extracting expressive molecular features is essential for molecular property prediction. Sequence-based representation is a common representation of molecules, which ignores the structure information of molecules. While molecular graph representation has a weak ability in expressing the 3D structure. In this article, we try to make use of the advantages of different type representations simultaneously for molecular property prediction. Thus, we propose a fusion model named DLF-MFF, which integrates the multi-type molecular features. Specifically, we first extract four different types of features from molecular fingerprints, 2D molecular graph, 3D molecular graph and molecular image. Then, in order to learn molecular features individually, we use four essential deep learning frameworks, which correspond to four distinct molecular representations. The final molecular representation is created by integrating the four feature vectors and feeding them into prediction layer to predict molecular property. We compare DLF-MFF with 7 state-of-the-art methods on 6 benchmark datasets consisting of multiple molecular properties, the experimental results show that DLF-MFF achieves state-of-the-art performance on 6 benchmark datasets. Moreover, DLF-MFF is applied to identify potential anti-SARS-CoV-2 inhibitor from 2500 drugs. We predict probability of each drug being inferred as a 3CL protease inhibitor and also calculate the binding affinity scores between each drug and 3CL protease. The results show that DLF-MFF product better performance in the identification of anti-SARS-CoV-2 inhibitor. This work is expected to offer novel research perspectives for accurate prediction of molecular properties and provide valuable insights into drug repurposing for COVID-19.

PmxPred: A data-driven approach for the identification of active polymyxin analogues against gram-negative bacteria

The multidrug-resistant Gram-negative bacteria has evolved into a worldwide threat to human health; over recent decades, polymyxins have re-emerged in clinical practice due to their high activity against multidrug-resistant bacteria. Nevertheless, the nephrotoxicity and neurotoxicity of polymyxins seriously hinder their practical use in the clinic. Based on the quantitative structure-activity relationship (QSAR), analogue design is an efficient strategy for discovering biologically active compounds with fewer adverse effects. To accelerate the polymyxin analogues discovery process and find the polymyxin analogues with high antimicrobial activity against Gram-negative bacteria, here we developed PmxPred, a GCN and catBoost-based machine learning framework. The RDKit descriptors were used for the molecule and residues representation, and the ensemble learning model was utilized for the antimicrobial activity prediction. This framework was trained and evaluated on multiple Gram-negative bacteria datasets, including Acinetobacter baumannii, Escherichia coli, Klebsiella pneumoniae, Pseudomonas aeruginosa and a general Gram-negative bacteria dataset achieving an AUROC of 0.857, 0.880, 0.756, 0.895 and 0.865 on the independent test, respectively. PmxPred outperformed the transfer learning method that trained on 10 million molecules. We interpreted our model well-trained model by analysing the importance of global and residue features. Overall, PmxPred provides a powerful additional tool for predicting active polymyxin analogues, and holds the potential elucidate the mechanisms underlying the antimicrobial activity of polymyxins. The source code is publicly available on GitHub (https://github.com/yanwu20/PmxPred).

TransFoxMol: predicting molecular property with focused attention

Predicting the biological properties of molecules is crucial in computer-aided drug development, yet it's often impeded by data scarcity and imbalance in many practical applications. Existing approaches are based on self-supervised learning or 3D data and using an increasing number of parameters to improve performance. These approaches may not take full advantage of established chemical knowledge and could inadvertently introduce noise into the respective model. In this study, we introduce a more elegant transformer-based framework with focused attention for molecular representation (TransFoxMol) to improve the understanding of artificial intelligence (AI) of molecular structure property relationships. TransFoxMol incorporates a multi-scale 2D molecular environment into a graph neural network + Transformer module and uses prior chemical maps to obtain a more focused attention landscape compared to that obtained using existing approaches. Experimental results show that TransFoxMol achieves state-of-the-art performance on MoleculeNet benchmarks and surpasses the performance of baselines that use self-supervised learning or geometry-enhanced strategies on small-scale datasets. Subsequent analyses indicate that TransFoxMol's predictions are highly interpretable and the clever use of chemical knowledge enables AI to perceive molecules in a simple but rational way, enhancing performance.

Modality-DTA: Multimodality Fusion Strategy for Drug-Target Affinity Prediction

Prediction of the drug-target affinity (DTA) plays an important role in drug discovery. Existing deep learning methods for DTA prediction typically leverage a single modality, namely simplified molecular input line entry specification (SMILES) or amino acid sequence to learn representations. SMILES or amino acid sequences can be encoded into different modalities. Multimodality data provide different kinds of information, with complementary roles for DTA prediction. We propose Modality-DTA, a novel deep learning method for DTA prediction that leverages the multimodality of drugs and targets. A group of backward propagation neural networks is applied to ensure the completeness of the reconstruction process from the latent feature representation to original multimodality data. The tag between the drug and target is used to reduce the noise information in the latent representation from multimodality data. Experiments on three benchmark datasets show that our Modality-DTA outperforms existing methods in all metrics. Modality-DTA reduces the mean square error by 15.7% and improves the area under the precisionrecall curve by 12.74% in the Davis dataset. We further find that the drug modality Morgan fingerprint and the target modality generated by one-hot-encoding play the most significant roles. To the best of our knowledge, Modality-DTA is the first method to explore multimodality for DTA prediction.

Pushing the Boundaries of Molecular Property Prediction for Drug Discovery with Multitask Learning BERT Enhanced by SMILES Enumeration

Accurate prediction of pharmacological properties of small molecules is becoming increasingly important in drug discovery. Traditional feature-engineering approaches heavily rely on handcrafted descriptors and/or fingerprints, which need extensive human expert knowledge. With the rapid progress of artificial intelligence technology, data-driven deep learning methods have shown unparalleled advantages over feature-engineering-based methods. However, existing deep learning methods usually suffer from the scarcity of labeled data and the inability to share information between different tasks when applied to predicting molecular properties, thus resulting in poor generalization capability. Here, we proposed a novel multitask learning BERT (Bidirectional Encoder Representations from Transformer) framework, named MTL-BERT, which leverages large-scale pre-training, multitask learning, and SMILES (simplified molecular input line entry specification) enumeration to alleviate the data scarcity problem. MTL-BERT first exploits a large amount of unlabeled data through self-supervised pretraining to mine the rich contextual information in SMILES strings and then fine-tunes the pretrained model for multiple downstream tasks simultaneously by leveraging their shared information. Meanwhile, SMILES enumeration is used as a data enhancement strategy during the pretraining, fine-tuning, and test phases to substantially increase data diversity and help to learn the key relevant patterns from complex SMILES strings. The experimental results showed that the pretrained MTL-BERT model with few additional fine-tuning can achieve much better performance than the state-of-the-art methods on most of the 60 practical molecular datasets. Additionally, the MTL-BERT model leverages attention mechanisms to focus on SMILES character features essential to target properties for model interpretability.

Deep learning joint models for extracting entities and relations in biomedical: a survey and comparison

The rapid development of biomedicine has produced a large number of biomedical written materials. These unstructured text data create serious challenges for biomedical researchers to find information. Biomedical named entity recognition (BioNER) and biomedical relation extraction (BioRE) are the two most fundamental tasks of biomedical text mining. Accurately and efficiently identifying entities and extracting relations have become very important. Methods that perform two tasks separately are called pipeline models, and they have shortcomings such as insufficient interaction, low extraction quality and easy redundancy. To overcome the above shortcomings, many deep learning-based joint name entity recognition and relation extraction models have been proposed, and they have achieved advanced performance. This paper comprehensively summarize deep learning models for joint name entity recognition and relation extraction for biomedicine. The joint BioNER and BioRE models are discussed in the light of the challenges existing in the BioNER and BioRE tasks. Five joint BioNER and BioRE models and one pipeline model are selected for comparative experiments on four biomedical public datasets, and the experimental results are analyzed. Finally, we discuss the opportunities for future development of deep learning-based joint BioNER and BioRE models.

iEnhancer-DLRA: identification of enhancers and their strengths by a self-attention fusion strategy for local and global features

Identification and classification of enhancers are highly significant because they play crucial roles in controlling gene transcription. Recently, several deep learning-based methods for identifying enhancers and their strengths have been developed. However, existing methods are usually limited because they use only local or only global features. The combination of local and global features is critical to further improve the prediction performance. In this work, we propose a novel deep learning-based method, called iEnhancer-DLRA, to identify enhancers and their strengths. iEnhancer-DLRA extracts local and multi-scale global features of sequences by using a residual convolutional network and two bidirectional long short-term memory networks. Then, a self-attention fusion strategy is proposed to deeply integrate these local and global features. The experimental results on the independent test dataset indicate that iEnhancer-DLRA performs better than nine existing state-of-the-art methods in both identification and classification of enhancers in almost all metrics. iEnhancer-DLRA achieves 13.8% (for identifying enhancers) and 12.6% (for classifying strengths) improvement in accuracy compared with the best existing state-of-the-art method. This is the first time that the accuracy of an enhancer identifier exceeds 0.9 and the accuracy of the enhancer classifier exceeds 0.8 on the independent test set. Moreover, iEnhancer-DLRA achieves superior predictive performance on the rice dataset compared with the state-of-the-art method RiceENN.

iPiDA-LTR: Identifying piwi-interacting RNA-disease associations based on Learning to Rank

Piwi-interacting RNAs (piRNAs) are regarded as drug targets and biomarkers for the diagnosis and therapy of diseases. However, biological experiments cost substantial time and resources, and the existing computational methods only focus on identifying missing associations between known piRNAs and diseases. With the fast development of biological experiments, more and more piRNAs are detected. Therefore, the identification of piRNA-disease associations of newly detected piRNAs has significant theoretical value and practical significance on pathogenesis of diseases. In this study, the iPiDA-LTR predictor is proposed to identify associations between piRNAs and diseases based on Learning to Rank. The iPiDA-LTR predictor not only identifies the missing associations between known piRNAs and diseases, but also detects diseases associated with newly detected piRNAs. Experimental results demonstrate that iPiDA-LTR effectively predicts piRNA-disease associations outperforming the other related methods.

A brief review of protein-ligand interaction prediction

The task of identifying protein–ligand interactions (PLIs) plays a prominent role in the field of drug discovery. However, it is infeasible to identify potential PLIs via costly and laborious in vitro experiments. There is a need to develop PLI computational prediction approaches to speed up the drug discovery process. In this review, we summarize a brief introduction to various computation-based PLIs. We discuss these approaches, in particular, machine learning-based methods, with illustrations of different emphases based on mainstream trends. Moreover, we analyzed three research dynamics that can be further explored in future studies.

ATGPred-FL: sequence-based prediction of autophagy proteins with feature representation learning

Autophagy plays an important role in biological evolution and is regulated by many autophagy proteins. Accurate identification of autophagy proteins is crucially important to reveal their biological functions. Due to the expense and labor cost of experimental methods, it is urgent to develop automated, accurate and reliable sequence-based computational tools to enable the identification of novel autophagy proteins among numerous proteins and peptides. For this purpose, a new predictor named ATGPred-FL was proposed for the efficient identification of autophagy proteins. We investigated various sequence-based feature descriptors and adopted the feature learning method to generate corresponding, more informative probability features. Then, a two-step feature selection strategy based on accuracy was utilized to remove irrelevant and redundant features, leading to the most discriminative 14-dimensional feature set. The final predictor was built using a support vector machine classifier, which performed favorably on both the training and testing sets with accuracy values of 94.40% and 90.50%, respectively. ATGPred-FL is the first ATG machine learning predictor based on protein primary sequences. We envision that ATGPred-FL will be an effective and useful tool for autophagy protein identification, and it is available for free at http://lab.malab.cn/~acy/ATGPred-FL , the source code and datasets are accessible at https://github.com/jiaoshihu/ATGPred .

Immunoglobulin Classification Based on FC* and GC* Features

Immunoglobulins have a pivotal role in disease regulation. Therefore, it is vital to accurately identify immunoglobulins to develop new drugs and research related diseases. Compared with utilizing high-dimension features to identify immunoglobulins, this research aimed to examine a method to classify immunoglobulins and non-immunoglobulins using two features, FC* and GC*. Classification of 228 samples (109 immunoglobulin samples and 119 non-immunoglobulin samples) revealed that the overall accuracy was 80.7% in 10-fold cross-validation using the J48 classifier implemented in Weka software. The FC* feature identified in this study was found in the immunoglobulin subtype domain, which demonstrated that this extracted feature could represent functional and structural properties of immunoglobulins for forecasting.

AAPred-CNN: accurate predictor based on deep convolution neural network for identification of anti-angiogenic peptides

Recently, deep learning techniques have been developed for various bioactive peptide prediction tasks. However, there are only conventional machine learning-based methods for the prediction of anti-angiogenic peptides (AAP), which play an important role in cancer treatment. The main reason why no deep learning method has been involved in this field is that there are too few experimentally validated AAPs to support the training of deep models but researchers have believed that deep learning seriously depends on the amounts of labeled data. In this paper, as a tentative work, we try to predict AAP by constructing different classical deep learning models and propose the first deep convolution neural network-based predictor (AAPred-CNN) for AAP. Contrary to intuition, the experimental results show that deep learning models can achieve superior or comparable performance to the state-of-the-art model, although they are given a few labeled sequences to train. We also decipher the influence of hyper-parameters and training samples on the performance of deep learning models to help understand how the model work. Furthermore, we also visualize the learned embeddings by dimension reduction to increase the model interpretability and reveal the residue propensity of AAP through the statistics of convolutional features for different residues. In summary, this work demonstrates the powerful representation ability of AAPred-CNNfor AAP prediction, further improving the prediction accuracy of AAP.

Learning spatial structures of proteins improves protein-protein interaction prediction

Spatial structures of proteins are closely related to protein functions. Integrating protein structures improves the performance of protein-protein interaction (PPI) prediction. However, the limited quantity of known protein structures restricts the application of structure-based prediction methods. Utilizing the predicted protein structure information is a promising method to improve the performance of sequence-based prediction methods. We propose a novel end-to-end framework, TAGPPI, to predict PPIs using protein sequence alone. TAGPPI extracts multi-dimensional features by employing 1D convolution operation on protein sequences and graph learning method on contact maps constructed from AlphaFold. A contact map contains abundant spatial structure information, which is difficult to obtain from 1D sequence data directly. We further demonstrate that the spatial information learned from contact maps improves the ability of TAGPPI in PPI prediction tasks. We compare the performance of TAGPPI with those of nine state-of-the-art sequence-based methods, and TAGPPI outperforms such methods in all metrics. To the best of our knowledge, this is the first method to use the predicted protein topology structure graph for sequence-based PPI prediction. More importantly, our proposed architecture could be extended to other prediction tasks related to proteins.