De-novo protein function prediction using DNA binding and RNA binding proteins as a test case

Identification of DNA-protein binding residues through integration of Transformer encoder and Bi-directional Long Short-Term Memory

DNA-protein binding is crucial for the normal development and function of organisms. The significance of accurately identifying DNA-protein binding sites lies in its role in disease prevention and the development of innovative approaches to disease treatment. In the present study, we introduce a precise and robust identifier for DNA-protein binding residues. In the context of protein representation, we combine the evolutionary information of the protein, represented by its position-specific scoring matrix, with the spatial information of the protein's secondary structure, enriching the overall informational content. This approach initially employs a combination of Bi-directional Long Short-Term Memory and Transformer encoder to jointly extract the interdependencies among residues within the protein sequence. Subsequently, convolutional operations are applied to the resulting feature matrix to capture local features of the residues. Experimental results on the benchmark dataset demonstrate that our method exhibits a higher level of competitiveness when compared to contemporary classifiers. Specifically, our method achieved an MCC of 0.349, SP of 96.50%, SN of 44.03% and ACC of 94.59% on the PDNA-41 dataset.

PredDRBP-MLP: Prediction of DNA-binding proteins and RNA-binding proteins by multilayer perceptron

Proteins interact with many molecules in order to maintain the vital activities in cells. Proteins that interact with DNA are called DNA-binding proteins (DBP), and proteins that interact with RNA are called RNA-binding proteins (RBP). Since DBPs and RBPs are involved in critical biological processes, their classification is quite important. Although the convolutional neural network and bidirectional long-short-term memory hybrid model (CNN-BiLSTM) is very popular in DBP and RBP classification, it has problems such as requirement of high processing power and long training time. Therefore, a multilayer perceptron (MLP) based predictor, PredDRBP-MLP (Predictor of DNA-Binding Proteins and RNA-Binding Proteins - Multilayer Perceptron) was developed in this study. PredDRBP-MLP is an artificial learning model that performs multi-class classification of DBPs, RBPs and non-nucleic acid-binding proteins (NNABP). PredDRBP-MLP achieved quite successful results on the independent dataset, specifically in the NNABP class, compared to the existing predictors, in addition to requiring lower processing power and being able to train quicker compared to CNN-BiLSTM based predictors. In NNABP class, PredDRBP-MLP predictor achieved 0.578 precision, 0.522 recall and 0.549 F1-score, while other multi-class predictor achieved 0.486 precision, 0.183 recall and 0.266 F1-score. A desktop application was developed for PredDRBP-MLP. The application is freely accessible at https://sourceforge.net/projects/preddrbp-mlp.

PAI-1 is a potential transcriptional silencer that supports bladder cancer cell activity

The extracellular activity of Plasminogen activator inhibitor-1 (PAI-1) is well described, acting as an inhibitor of tissue plasminogen activator and urokinase-type plasminogen activator, impacting fibrinolysis. Recent studies have revealed a pro-tumorigenic role of PAI-1 in human cancers, via the regulation of angiogenesis and tumor cell survival. In this study, immunohistochemical staining of 939 human bladder cancer specimens showed that PAI-1 expression levels correlated with tumor grade, tumor stage and overall survival. The typical subcellular localization of PAI-1 is cytoplasmic, but in approximately a quarter of the cases, PAI-1 was observed to be localized to both the tumor cell cytoplasm and the nucleus. To investigate the potential function of nuclear PAI-1 in tumor biology we applied chromatin immunoprecipitation (ChIP)-sequencing, gene expression profiling, and rapid immunoprecipitation mass spectrometry to a pair of bladder cancer cell lines. ChIP-sequencing revealed that PAI-1 can bind DNA at distal intergenic regions, suggesting a role as a transcriptional coregulator. The downregulation of PAI-1 in bladder cancer cell lines caused the upregulation of numerous genes, and the integration of ChIP-sequence and RNA-sequence data identified 57 candidate genes subject to PAI-1 regulation. Taken together, the data suggest that nuclear PAI-1 can influence gene expression programs and support malignancy.

In silico approach to probe the binding affinity between OMVs harboring the ZEGFR affibody and the EGF receptor

There is a growing interest in designing a nanocarrier containing an EGFR targeting affibody to direct toward cancer cells. Here, cytolysin A was cloned at the N-terminus of Z_EGFR:1907 affibody to guarantee its surface presentation on the OMVs while targeting the epidermal growth factor receptors (EGFRs). A separate construct including a fusogenic peptide (GALA) was also designed for the endosomal escape of the nanocarrier. Binding of the two constructs ClyA-affi_EGFR and ClyA-affi_EGFR-GALA to domain III of EGFR was investigated using molecular docking and molecular dynamic simulations. The higher stability of the ClyA-affi_EGFR-GALA/EGFR as compared to the ClyA-affi_EGFR/EGFR complex was evident. The ClyA-affi_EGFR-GALA structure showed a higher RMSD during the first half of the simulation time implying a much less stable behavior. Plateau state of the radius of gyration plot of ClyA-affi_EGFR-GALA confirmed a well-folded structure in the presence of the GALA sequence. Solvent accessible surface area for both proteins was in the same range. The data obtained from hydrogen bond analysis revealed a more equilibrated and stable form of the ClyA-affi_EGFR-GALA structure upon interaction with EGFR. The data provided here was a requisite for our biological evaluation of the synthesized constructs as a component of a novel drug delivery system.

Identifying protein function and functional links based on large-scale co-occurrence patterns

Objective

The vast majority of known proteins have not been experimentally tested even at the level of measuring their expression, and the function of many proteins remains unknown. In order to decipher protein function and examine functional associations, we developed "Cliquely", a software tool based on the exploration of co-occurrence patterns.

Computational model

Using a set of more than 23 million proteins divided into 404,947 orthologous clusters, we explored the co-occurrence graph of 4,742 fully sequenced genomes from the three domains of life. Edge weights in this graph represent co-occurrence probabilities. We use the Bron–Kerbosch algorithm to detect maximal cliques in this graph, fully-connected subgraphs that represent meaningful biological networks from different functional categories.

Main results

We demonstrate that Cliquely can successfully identify known networks from various pathways, including nitrogen fixation, glycolysis, methanogenesis, mevalonate and ribosome proteins. Identifying the virulence-associated type III secretion system (T3SS) network, Cliquely also added 13 previously uncharacterized novel proteins to the T3SS network, demonstrating the strength of this approach. Cliquely is freely available and open source. Users can employ the tool to explore co-occurrence networks using a protein of interest and a customizable level of stringency, either for the entire dataset or for a one of the three domains—Archaea, Bacteria, or Eukarya.

DeepMC-iNABP: Deep learning for multiclass identification and classification of nucleic acid-binding proteins

Nucleic acid-binding proteins (NABPs), including DNA-binding proteins (DBPs) and RNA-binding proteins (RBPs), play vital roles in gene expression. Accurate identification of these proteins is crucial. However, there are two existing challenges: one is the problem of ignoring DNA- and RNA-binding proteins (DRBPs), and the other is a cross-predicting problem referring to DBP predictors predicting DBPs as RBPs, and vice versa. In this study, we proposed a computational predictor, called DeepMC-iNABP, with the goal of solving these difficulties by utilizing a multiclass classification strategy and deep learning approaches. DBPs, RBPs, DRBPs and non-NABPs as separate classes of data were used for training the DeepMC-iNABP model. The results on test data collected in this study and two independent test datasets showed that DeepMC-iNABP has a strong advantage in identifying the DRBPs and has the ability to alleviate the cross-prediction problem to a certain extent. The web-server of DeepMC-iNABP is freely available at http://www.deepmc-inabp.net/. The datasets used in this research can also be downloaded from the website.

Improved datasets and evaluation methods for the automatic prediction of DNA-binding proteins

MOTIVATION

Accurate automatic annotation of protein function relies on both innovative models and robust datasets. Due to their importance in biological processes, the identification of DNA-binding proteins directly from protein sequence has been the focus of many studies. However, the datasets used to train and evaluate these methods have suffered from substantial flaws. We describe some of the weaknesses of the datasets used in previous DNA-binding protein literature and provide several new datasets addressing these problems. We suggest new evaluative benchmark tasks that more realistically assess real-world performance for protein annotation models. We propose a simple new model for the prediction of DNA-binding proteins and compare its performance on the improved datasets to two previously published models. In addition, we provide extensive tests showing how the best models predict across taxa.

RESULTS

Our new gradient boosting model, which uses features derived from a published protein language model, outperforms the earlier models. Perhaps surprisingly, so does a baseline nearest neighbor model using BLAST percent identity. We evaluate the sensitivity of these models to perturbations of DNA-binding regions and control regions of protein sequences. The successful data-driven models learn to focus on DNA-binding regions. When predicting across taxa, the best models are highly accurate across species in the same kingdom and can provide some information when predicting across kingdoms.

AVAILABILITY AND IMPLEMENTATION

The data and results for this article can be found at https://doi.org/10.5281/zenodo.5153906. The code for this article can be found at https://doi.org/10.5281/zenodo.5153683. The code, data and results can also be found at https://github.com/AZaitzeff/tools_for_dna_binding_proteins.

Two novel RNA-binding proteins identification through computational prediction and experimental validation

Since RBPs play important roles in the cell, it's particularly important to find new RBPs. We performed iRIP-seq and CLIP-seq to verify two proteins, CLIP1 and DMD, predicted by RBPPred whether are RBPs or not. The experimental results confirm that these two proteins have RNA-binding activity. We identified significantly enriched binding motifs UGGGGAGG, CUUCCG and CCCGU for CLIP1 (iRIP-seq), DMD (iRIP-seq) and DMD (CLIP-seq), respectively. The computational KEGG and GO analysis show that the CLIP1 and DMD share some biological processes and functions. Besides, we found that the SNPs between DMD and its RNA partners may be associated with Becker muscular dystrophy, Duchenne muscular dystrophy, Dilated cardiomyopathy 3B and Cardiovascular phenotype. Among the thirteen cancers data, CLIP1 and another 300 oncogenes always co-occur, and 123 of these 300 genes interact with CLIP1. These cancers may be associated with the mutations occurred in both CLIP1 and the genes it interacts with.

dSPRINT: predicting DNA, RNA, ion, peptide and small molecule interaction sites within protein domains

Domains are instrumental in facilitating protein interactions with DNA, RNA, small molecules, ions and peptides. Identifying ligand-binding domains within sequences is a critical step in protein function annotation, and the ligand-binding properties of proteins are frequently analyzed based upon whether they contain one of these domains. To date, however, knowledge of whether and how protein domains interact with ligands has been limited to domains that have been observed in co-crystal structures; this leaves approximately two-thirds of human protein domain families uncharacterized with respect to whether and how they bind DNA, RNA, small molecules, ions and peptides. To fill this gap, we introduce dSPRINT, a novel ensemble machine learning method for predicting whether a domain binds DNA, RNA, small molecules, ions or peptides, along with the positions within it that participate in these types of interactions. In stringent cross-validation testing, we demonstrate that dSPRINT has an excellent performance in uncovering ligand-binding positions and domains. We also apply dSPRINT to newly characterize the molecular functions of domains of unknown function. dSPRINT's predictions can be transferred from domains to sequences, enabling predictions about the ligand-binding properties of 95% of human genes. The dSPRINT framework and its predictions for 6503 human protein domains are freely available at http://protdomain.princeton.edu/dsprint.

Refinement of bamboo genome annotations through integrative analyses of transcriptomic and epigenomic data

Bamboo, one of the most crucial nontimber forest resources worldwide, has the capacity for rapid growth. In recent years, the genome of moso bamboo (Phyllostachys edulis) has been decoded, and a large amount of transcriptome data has been published. In this study, we generated the genome-wide profiles of the histone modification H3K4me3 in leaf, stem, and root tissues of bamboo. The trends in the distribution patterns were similar to those in rice. We developed a processing pipeline for predicting novel transcripts to refine the structural annotation of the genome using H3K4me3 ChIP-seq data and 29 RNA-seq datasets. As a result, 12,460 novel transcripts were predicted in the bamboo genome. Compared with the transcripts in the newly released version 2.0 of the bamboo genome, these novel transcripts are tissue-specific and shorter, and most have a single exon. Some representative novel transcripts were validated by semiquantitative RT-PCR and qRT-PCR analyses. Furthermore, we put these novel transcripts back into the ChIP-seq analysis pipeline and discovered that the percentages of H3K4me3 in genic elements were increased. Overall, this work integrated transcriptomic data and epigenomic data to refine the annotation of the genome in order to discover more functional genes and study bamboo growth and development, and the application of this predicted pipeline may help refine the structural annotation of the genome in other species.

Identity: rapid alignment-free prediction of sequence alignment identity scores using self-supervised general linear models

Pairwise global alignment is a fundamental step in sequence analysis. Optimal alignment algorithms are quadratic-slow especially on long sequences. In many applications that involve large sequence datasets, all what is needed is calculating the identity scores (percentage of identical nucleotides in an optimal alignment-including gaps-of two sequences); there is no need for visualizing how every two sequences are aligned. For these applications, we propose Identity, which produces global identity scores for a large number of pairs of DNA sequences using alignment-free methods and self-supervised general linear models. For the first time, the new tool can predict pairwise identity scores in linear time and space. On two large-scale sequence databases, Identity provided the best compromise between sensitivity and precision while being faster than BLAST, Mash, MUMmer4 and USEARCH by 2-80 times. Identity was the best performing tool when searching for low-identity matches. While constructing phylogenetic trees from about 6000 transcripts, the tree due to the scores reported by Identity was the closest to the reference tree (in contrast to andi, FSWM and Mash). Identity is capable of producing pairwise identity scores of millions-of-nucleotides-long bacterial genomes; this task cannot be accomplished by any global-alignment-based tool. Availability: https://github.com/BioinformaticsToolsmith/Identity.

Long noncoding RNA FGF14-AS2 inhibits breast cancer metastasis by regulating the miR-370-3p/FGF14 axis

Long noncoding RNAs (lncRNAs) have emerged as important regulators in cancers, including breast cancer. However, the overall biological roles and clinical significance of most lncRNAs are not fully understood. This study aimed to elucidate the potential role of a novel lncRNA FGF14-AS2 and the mechanisms underlying metastasis in breast cancer. The lncRNA FGF14-AS2 was significantly downregulated in breast cancer tissues; patients with lower FGF14-AS2 expression had advanced clinical stage. In vitro and in vivo assays of FGF14-AS2 alterations revealed a complex integrated phenotype affecting breast cancer cell migration, invasion, and tumor metastasis. Mechanistically, FGF14-AS2 functioned as a competing endogenous RNA of miR-370-3p, thereby leading to the activation of its coding counterpart, FGF14. Clinically, we observed increased miR-370-3p expression in breast cancer tissues, whereas FGF14 expression was decreased in breast cancer tissues compared to the adjacent normal breast tissues. FGF14-AS2 expression was significantly negatively correlated with miR-370-3p expression, and correlated positively to FGF14 expression. Collectively, our findings support a model in which the FGF14-AS2/miR-370-3p/FGF14 axis is a critical regulator in breast cancer metastasis, suggesting a new therapeutic direction in breast cancer.

Fibroblast Growth Factor-14 Acts as Tumor Suppressor in Lung Adenocarcinomas

Investigation of the molecular dynamics in lung cancer is crucial for the development of new treatment strategies. Fibroblast growth factor (FGF) 14 belongs to the FGF family, which might play a crucial role in cancer progression. We analyzed lung adenocarcinoma (LUAC) patients samples and found that FGF14 was downregulated, correlating with reduced survival and oncogenic mutation status. FGF14 overexpression in lung cancer cell lines resulted in decreased proliferation, colony formation, and migration, as well as increased expression of epithelial markers and a decreased expression of mesenchymal markers, indicating a mesenchymal to epithelial transition in vitro. We verified these findings using small interfering RNA against FGF14 and further confirmed the suppressive effect of FGF14 in a NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ immunodeficient xenograft tumor model. Moreover, FGF14 overexpressing tumor cell RNA sequencing data suggests that genes affected by FGF14 were related to the extracellular matrix, playing a role in proliferation and migration. Notably, newly identified FGF14 target genes, adenosine deaminase RNA specific B1 (ADARB1), collagen and calcium-binding epidermal growth factor domain-containing protein 1 (CCBE1), α1 chain of collagen XI (COL11A1), and mucin 16 (MUC16) expression was negatively correlated with overall survival when FGF14 was downregulated in LUAC. These findings led us to suggest that FGF14 regulates proliferation and migration in LUAC.

FGF14 Functions as a Tumor Suppressor through Inhibiting PI3K/AKT/mTOR Pathway in Colorectal Cancer

We identified that Fibroblast Growth Factor 14 (FGF14) was preferentially methylated in colorectal cancer (CRC). In this study, we aimed to investigate the epigenetic regulation, biological function and molecular mechanism of FGF14 in CRC. The expression of FGF14 in CRC cell lines, normal human colon epithelial cell line, CRC tissues and paired adjacent normal tissues was detected by PCR and Western blot. The biological function of FGF14 in CRC was interrogated by cell viability assay, colony formation, flow cytometry, cell invasion and migration assay, as well as in vivo study. We found FGF14 was downregulated or silenced in all (10/10) CRC cell lines, while it was expressed in normal colonic tissues and normal human colon epithelial cell line. The expression of FGF14 was lower in primary CRCs as compared to their adjacent normal tissues. Significant higher methylation of FGF14 was observed in CRCs than that in normal tissues based on the data from TCGA database. The loss of FGF14 gene expression was restored by treatment with DNA methyltransferase inhibitor 5-Aza. Re-expression of FGF14 in CRC cell lines inhibited cell viability and colony formation, and induced cell apoptosis. FGF14 induced mitochondrial apoptosis and inhibited PI3K/AKT/mTOR pathway. In xenograft mouse model, overexpression of FGF14 significantly reduced tumor growth (P<0.001). In conclusion, FGF14 is a novel tumor suppressor, which suppresses cell proliferation and induces cell apoptosis via mediating PI3K/AKT/mTOR pathway.

Refinement of evolutionary medicine predictions based on clinical evidence for the manifestations of Mendelian diseases

Prediction methods have become an integral part of biomedical and biotechnological research. However, their clinical interpretations are largely based on biochemical or molecular data, but not clinical data. Here, we focus on improving the reliability and clinical applicability of prediction algorithms. We assembled and curated two large non-overlapping large databases of clinical phenotypes. These phenotypes were caused by missense variations in 44 and 63 genes associated with Mendelian diseases. We used these databases to establish and validate the model, allowing us to improve the predictions obtained from EVmutation, SNAP2 and PoPMuSiC 2.1. The predictions of clinical effects suffered from a lack of specificity, which appears to be the common constraint of all recently used prediction methods, although predictions mediated by these methods are associated with nearly absolute sensitivity. We introduced evidence-based tailoring of the default settings of the prediction methods; this tailoring substantially improved the prediction outcomes. Additionally, the comparisons of the clinically observed and theoretical variations led to the identification of large previously unreported pools of variations that were under negative selection during molecular evolution. The evolutionary variation analysis approach described here is the first to enable the highly specific identification of likely disease-causing missense variations that have not yet been associated with any clinical phenotype.

Machine learning techniques for protein function prediction

Proteins play important roles in living organisms, and their function is directly linked with their structure. Due to the growing gap between the number of proteins being discovered and their functional characterization (in particular as a result of experimental limitations), reliable prediction of protein function through computational means has become crucial. This paper reviews the machine learning techniques used in the literature, following their evolution from simple algorithms such as logistic regression to more advanced methods like support vector machines and modern deep neural networks. Hyperparameter optimization methods adopted to boost prediction performance are presented. In parallel, the metamorphosis in the features used by these algorithms from classical physicochemical properties and amino acid composition, up to text-derived features from biomedical literature and learned feature representations using autoencoders, together with feature selection and dimensionality reduction techniques, are also reviewed. The success stories in the application of these techniques to both general and specific protein function prediction are discussed.

Protein functional annotation of simultaneously improved stability, accuracy and false discovery rate achieved by a sequence-based deep learning

Functional annotation of protein sequence with high accuracy has become one of the most important issues in modern biomedical studies, and computational approaches of significantly accelerated analysis process and enhanced accuracy are greatly desired. Although a variety of methods have been developed to elevate protein annotation accuracy, their ability in controlling false annotation rates remains either limited or not systematically evaluated. In this study, a protein encoding strategy, together with a deep learning algorithm, was proposed to control the false discovery rate in protein function annotation, and its performances were systematically compared with that of the traditional similarity-based and de novo approaches. Based on a comprehensive assessment from multiple perspectives, the proposed strategy and algorithm were found to perform better in both prediction stability and annotation accuracy compared with other de novo methods. Moreover, an in-depth assessment revealed that it possessed an improved capacity of controlling the false discovery rate compared with traditional methods. All in all, this study not only provided a comprehensive analysis on the performances of the newly proposed strategy but also provided a tool for the researcher in the fields of protein function annotation.