Predicted Binding Sites Research Articles

AD‐Associated Variants Directly Impair Surface Interactions Between the Hydrophobic Site of TREM2 and the Hinge Region of ApoE

AbstractBackgroundTriggering Receptor Expressed on Myeloid Cells 2 (TREM2) and Apolipoprotein E (ApoE) are two of the strongest genetic risk factors for late‐onset Alzheimer’s disease (AD), but while their interplay in the microglial response to AD pathology is hotly studied, surprisingly little is known about the mechanism by which their variants impair this role.MethodTo define the structural mechanism by which AD‐associated variants impair the TREM2‐ApoE interaction, we used sequence‐based in silico methods to predict the binding site between TREM2 and ApoE, validated the predicted binding site with biolayer interferometry (BLI), then constructed initial models of TREM2‐ApoE complex structures using the identified binding site between TREM2 and APOE as restraints for unbiased molecular docking methods. The top four initial models were re‐ranked by comparing the effect of key residue mutations in the TREM2‐ApoE complex on binding affinities calculated from molecular dynamics‐molecular mechanics Poisson‐Boltzmann surface area (MD/MM‐PBSA) binding free energy analyses with binding affinities measured by BLI. The final validated TREM2‐ApoE complex structure model provides for further investigating the molecular and structural effects of AD‐associated TREM2 variants on interactions between hydrophobic site of TREM2 and hinge region of ApoE.ResultOur iterative MD/MM‐PBSA–BLI procedure generated an all‐atom model of the TREM2‐ApoE complex that replicates the binding effects of both engineered mutations in TREM2 and AD‐associated variants in both TREM2 and ApoE. The validated model shows the interaction centered on the apical hydrophobic site of TREM2 and the hinge region of ApoE. The interaction sites in the TREM2‐ApoE complex model are consistent with regions affected by AD‐associated variants in simulations of the individual proteins, with biophysical studies using truncated constructs of the two proteins, and with a recently reported minimal active peptide of TREM2 that is sufficient to stimulate its cytoprotective and cytokine regulating roles in primary microglia.ConclusionIterative molecular modeling and biophysical binding affinity assays show the hydrophobic site of TREM2 binding the hinge region of ApoE, an interaction which is impaired by AD‐associated variants and which can be targeted to generate novel therapeutics for AD.

Transcriptome Co-Expression Network Analysis of Peach Fruit with Different Sugar Concentrations Reveals Key Regulators in Sugar Metabolism Involved in Cold Tolerance.

Peach fruits are known to be highly susceptible to chilling injury (CI) during low-temperature storage, which has been linked to the level of sugar concentration in the fruit. In order to better understand the relationship between sugar metabolism and CI, we conducted a study examining the concentration of sucrose, fructose, and glucose in peach fruit with different sugar concentrations and examined their relationship with CI. Through transcriptome sequencing, we screened the functional genes and transcription factors (TFs) involved in the sugar metabolism pathway that may cause CI in peach fruit. Our results identified five key functional genes (PpSS, PpINV, PpMGAM, PpFRK, and PpHXK) and eight TFs (PpMYB1/3, PpMYB-related1, PpWRKY4, PpbZIP1/2/3, and PpbHLH2) that are associated with sugar metabolism and CI development. The analysis of co-expression network mapping and binding site prediction identified the most likely associations between these TFs and functional genes. This study provides insights into the metabolic and molecular mechanisms regulating sugar changes in peach fruit with different sugar concentrations and presents potential targets for breeding high-sugar and cold-tolerant peach varieties.

CRBP-HFEF: Prediction of RBP-Binding Sites on circRNAs Based on Hierarchical Feature Expansion and Fusion.

Circular RNAs (circRNAs) participate in the regulation of biological processes by binding to specific proteins and thus influence transcriptional processes. In recent years, circRNAs have become an emerging hotspot in RNA research. Due to powerful learning ability, the various deep learning frameworks have been used to predict the binding sites of RNA-binding protein (RPB) on circRNAs. These methods usually perform only single-level feature extraction of sequence information. However, the feature acquisition may be inadequate for single-level extraction. Generally, the features of deep and shallow layers of neural network can complement each other and are both important for binding site prediction tasks. Based on this concept, we propose a method that combines deep and shallow features, namely CRBP-HFEF. Specifically, features are first extracted and expanded for different levels of network. Then, the expanded deep and shallow features are fused and fed into the classification network, which finally determines whether they are binding sites. Compared to several existing methods, the experimental results on multiple datasets show that the proposed method achieves significant improvements in a number of metrics (with an average AUC of 0.9855). Moreover, much sufficient ablation experiments are also performed to verify the effectiveness of the hierarchical feature expansion strategy.

Argentatin C Analogues with Potential Antinociceptive Activity and Other Triterpenoid Constituents from the Aerial Parts of Parthenium incanum.

Four new triterpenes, 25-dehydroxy-25-methoxyargentatin C (1), 20S-hydroxyargentatin C (2), 20S-hydroxyisoargentatin C (3), and 24-epi-argentatin C (4), together with 10 known triterpenes (5-14) were isolated from the aerial parts of Parthenium incanum. The structures of 1-4 were elucidated by detailed analysis of their spectroscopic data, and the known compounds 5-14 were identified by comparison of their spectroscopic data with those reported. Since argentatin C (11) was found to exhibit antinociceptive activity by decreasing the excitability of rat and macaque dorsal root ganglia (DRG) neurons, 11 and its new analogues 1-4 were evaluated for their ability to decrease the excitability of rat DRG neurons. Of the argentatin C analogues tested, 25-dehydroxy-25-methoxyargentatin C (1) and 24-epi-argentatin C (4) decreased neuronal excitability in a manner comparable to 11. Preliminary structure-activity relationships for the action potential-reducing effects of argentatin C (11) and its analogues 1-4, and their predicted binding sites in pain-relevant voltage-gated sodium and calcium channels (VGSCs and VGCCs) in DRG neurons are presented.

DeepBindPPI: Protein-Protein Binding Site Prediction Using Attention Based Graph Convolutional Network.

Due to the importance of protein-protein interactions in defence mechanism of living body, attempts were made to investigate its attributes, including, but not limited to, binding affinity, and binding region. Contemporary strategies for binding site prediction largely resort to deep learning techniques but turned out to be low precision models. As laboratory experiments for drug discovery tasks utilize this information, increased false positives devalue the computational methods. This emphasize the need to develop enhanced strategies. DeepBindPPI employs deep learning technique to predict the binding regions of proteins, particularly antigen-antibody interaction sites. The results obtained are applied in a docking environment to confirm their correctness. An integration of graph convolutional network with attention mechanism predicts interacting amino acids with improved precision. The model learns the determining factors in interaction from a general pool of proteins and is then fine-tuned using antigen-antibody data. Comparison of the proposed method with existing techniques shows that the developed model has comparable performance. The use of a separate spatial network clearly improved the precision of the proposed method from 0.4 to 0.5. An attempt to utilize the interface information for docking using the HDOCK server gives promising results, with high-quality structures appearing in the top10 ranks.

Binding Site Detection Remastered: Enabling Fast, Robust, and Reliable Binding Site Detection and Descriptor Calculation with DoGSite3

Binding site prediction on protein structures is a crucial step in early phase drug discovery whenever experimental or predicted structure models are involved. DoGSite belongs to the widely used tools for this task. It is a grid-based method that uses a Difference-of-Gaussian filter to detect cavities on the protein surface. We recently reimplemented the first version of this method, released in 2010, focusing on improved binding site detection in the presence of ligands and optimized parameters for more robust, reliable, and fast predictions and binding site descriptor calculations. Here, we introduce the new version, DoGSite3, compare it to its predecessor, and re-evaluate DoGSite on published data sets for a large-scale comparative performance evaluation.

Prediction of cooperative homeodomain DNA binding sites from high-throughput-SELEX data.

Homeodomain proteins constitute one of the largest families of metazoan transcription factors. Genetic studies have demonstrated that homeodomain proteins regulate many developmental processes. Yet, biochemical data reveal that most bind highly similar DNA sequences. Defining how homeodomain proteins achieve DNA binding specificity has therefore been a long-standing goal. Here, we developed a novel computational approach to predict cooperative dimeric binding of homeodomain proteins using High-Throughput (HT) SELEX data. Importantly, we found that 15 of 88 homeodomain factors form cooperative homodimer complexes on DNA sites with precise spacing requirements. Approximately one third of the paired-like homeodomain proteins cooperatively bind palindromic sequences spaced 3 bp apart, whereas other homeodomain proteins cooperatively bind sites with distinct orientation and spacing requirements. Combining structural models of a paired-like factor with our cooperativity predictions identified key amino acid differences that help differentiate between cooperative and non-cooperative factors. Finally, we confirmed predicted cooperative dimer sites in vivo using available genomic data for a subset of factors. These findings demonstrate how HT-SELEX data can be computationally mined to predict cooperativity. In addition, the binding site spacing requirements of select homeodomain proteins provide a mechanism by which seemingly similar AT-rich DNA sequences can preferentially recruit specific homeodomain factors.

Regulation of VEGFR2 and AKT Signaling by Musashi-2 in Lung Cancer.

Lung cancer is the most frequently diagnosed cancer type and the leading cause of cancer-related deaths worldwide. Non-small cell lung cancer (NSCLC) represents most of the diagnoses of lung cancer. Vascular endothelial growth factor receptor-2 (VEGFR2) is a member of the VEGF family of receptor tyrosine kinase proteins, which are expressed on both endothelial and tumor cells, are one of the key proteins contributing to cancer development, and are involved in drug resistance. We previously showed that Musashi-2 (MSI2) RNA-binding protein is associated with NSCLC progression by regulating several signaling pathways relevant to NSCLC. In this study, we performed Reverse Protein Phase Array (RPPA) analysis of murine lung cancer, which suggests that VEGFR2 protein is strongly positively regulated by MSI2. Next, we validated VEGFR2 protein regulation by MSI2 in several human lung adenocarcinoma cell line models. Additionally, we found that MSI2 affected AKT signaling via negative PTEN mRNA translation regulation. In silico prediction analysis suggested that both VEGFR2 and PTEN mRNAs have predicted binding sites for MSI2. We next performed RNA immunoprecipitation coupled with quantitative PCR, which confirmed that MSI2 directly binds to VEGFR2 and PTEN mRNAs, suggesting a direct regulation mechanism. Finally, MSI2 expression positively correlated with VEGFR2 and VEGF-A protein levels in human lung adenocarcinoma samples. We conclude that the MSI2/VEGFR2 axis contributes to lung adenocarcinoma progression and is worth further investigations and therapeutic targeting.

Improving the Modeling of Extracellular Ligand Binding Pockets in RosettaGPCR for Conformational Selection.

G protein-coupled receptors (GPCRs) are the largest class of drug targets and undergo substantial conformational changes in response to ligand binding. Despite recent progress in GPCR structure determination, static snapshots fail to reflect the conformational space of putative binding pocket geometries to which small molecule ligands can bind. In comparative modeling of GPCRs in the absence of a ligand, often a shrinking of the orthosteric binding pocket is observed. However, the exact prediction of the flexible orthosteric binding site is crucial for adequate structure-based drug discovery. In order to improve ligand docking and guide virtual screening experiments in computer-aided drug discovery, we developed RosettaGPCRPocketSize. The algorithm creates a conformational ensemble of biophysically realistic conformations of the GPCR binding pocket between the TM bundle, which is consistent with a knowledge base of expected pocket geometries. Specifically, tetrahedral volume restraints are defined based on information about critical residues in the orthosteric binding site and their experimentally observed range of Cα-Cα-distances. The output of RosettaGPCRPocketSize is an ensemble of binding pocket geometries that are filtered by energy to ensure biophysically probable arrangements, which can be used for docking simulations. In a benchmark set, pocket shrinkage observed in the default RosettaGPCR was reduced by up to 80% and the binding pocket volume range and geometric diversity were increased. Compared to models from four different GPCR homology model databases (RosettaGPCR, GPCR-Tasser, GPCR-SSFE, and GPCRdb), the here-created models showed more accurate volumes of the orthosteric pocket when evaluated with respect to the crystallographic reference structure. Furthermore, RosettaGPCRPocketSize was able to generate an improved realistic pocket distribution. However, while being superior to other homology models, the accuracy of generated model pockets was comparable to AlphaFold2 models. Furthermore, in a docking benchmark using small-molecule ligands with a higher molecular weight between 400 and 700 Da, a higher success rate in creating native-like binding poses was observed. In summary, RosettaGPCRPocketSize can generate GPCR models with realistic orthosteric pocket volumes, which are useful for structure-based drug discovery applications.

From Deep Mutational Mapping of Allosteric Protein Landscapes to Deep Learning of Allostery and Hidden Allosteric Sites: Zooming in on "Allosteric Intersection" of Biochemical and Big Data Approaches.

The recent advances in artificial intelligence (AI) and machine learning have driven the design of new expert systems and automated workflows that are able to model complex chemical and biological phenomena. In recent years, machine learning approaches have been developed and actively deployed to facilitate computational and experimental studies of protein dynamics and allosteric mechanisms. In this review, we discuss in detail new developments along two major directions of allosteric research through the lens of data-intensive biochemical approaches and AI-based computational methods. Despite considerable progress in applications of AI methods for protein structure and dynamics studies, the intersection between allosteric regulation, the emerging structural biology technologies and AI approaches remains largely unexplored, calling for the development of AI-augmented integrative structural biology. In this review, we focus on the latest remarkable progress in deep high-throughput mining and comprehensive mapping of allosteric protein landscapes and allosteric regulatory mechanisms as well as on the new developments in AI methods for prediction and characterization of allosteric binding sites on the proteome level. We also discuss new AI-augmented structural biology approaches that expand our knowledge of the universe of protein dynamics and allostery. We conclude with an outlook and highlight the importance of developing an open science infrastructure for machine learning studies of allosteric regulation and validation of computational approaches using integrative studies of allosteric mechanisms. The development of community-accessible tools that uniquely leverage the existing experimental and simulation knowledgebase to enable interrogation of the allosteric functions can provide a much-needed boost to further innovation and integration of experimental and computational technologies empowered by booming AI field.

MBS: a genome browser annotation track for high-confident microRNA binding sites in whole human transcriptome.

MicroRNAs (miRNAs) are small non-coding ribonucleic acids (RNAs) that play a role in many regulatory pathways in eukaryotes. They usually exert their functions by binding mature messenger RNAs. The prediction of the binding targets of the endogenous miRNAs is crucial to unravel the processes they are involved in. In this work, we performed an extensive miRNA binding sites (MBS) prediction over all the annotated transcript sequences and made them available through an UCSC track. MBS annotation track allows to study and visualize the human miRNA binding sites transcriptome-wide in a genome browser, together with any other available information the user is interested in. In the creation of the database that underlies the MBS track, three consolidated algorithms of miRNA binding prediction have been used: PITA, miRanda and TargetScan, and information about the binding sites predicted by all of them has been collected. MBS track displays high-confident miRNA binding sites for the whole length of each human transcript, both coding and non-coding ones. Each annotation can redirect to a web page with the details of the miRNA binding and the involved transcripts. MBS can be easily applied to retrieve specific information such as the effects of alternative splicing on miRNA binding or when a specific miRNA binds an exon-exon junction in the mature RNA. Overall, MBS will be of great help for studying and visualizing, in a user-friendly mode, the predicted miRNA binding sites on all the transcripts arising from a gene or a region of interest. Database URL https://datasharingada.fondazionerimed.com:8080/MBS.

GeoBind: segmentation of nucleic acid binding interface on protein surface with geometric deep learning.

Unveiling the nucleic acid binding sites of a protein helps reveal its regulatory functions in vivo. Current methods encode protein sites from the handcrafted features of their local neighbors and recognize them via a classification, which are limited in expressive ability. Here, we present GeoBind, a geometric deep learning method for predicting nucleic binding sites on protein surface in a segmentation manner. GeoBind takes the whole point clouds of protein surface as input and learns the high-level representation based on the aggregation of their neighbors in local reference frames. Testing GeoBind on benchmark datasets, we demonstrate GeoBind is superior to state-of-the-art predictors. Specific case studies are performed to show the powerful ability of GeoBind to explore molecular surfaces when deciphering proteins with multimer formation. To show the versatility of GeoBind, we further extend GeoBind to five other types of ligand binding sites prediction tasks and achieve competitive performances.

Abstract 2024: High resolution limited proteolysis (HR-LiP), a novel structural proteomics approach for the prediction of small molecule-protein binding events

Abstract High resolution profiling of drug-protein interactions and binding mechanisms remains a major hurdle during lead selection and optimization. A key milestone in structure-based drug design is compound binding site identification and characterization. Structure-activity relationship (SAR) studies traditionally utilize time and cost intensive techniques such as nuclear magnetic resonance (NMR), x-ray crystallography (X-ray) and cryo-electron microscopy (cryo-EM). To date, hydrogen-deuterium exchange (HDX) is the only mass spectrometry-based approach that has been extensively utilized. Further, SAR studies are often complicated by protein size (large proteins and/or oligomers) and location (membrane proteins), which can lead to protocol adaptations that can introduce artifacts. High resolution limited proteolysis (HR-LiP) is a high-throughput, high-resolution approach based on LiP-Quant that utilizes peptide-level resolution to characterize drug-protein interactions. Two well-characterized drug target proteins, bromodomain-containing protein 4 (BRD4) and transitional endoplasmic reticulum ATPase (VCP), were selected for analysis as they represent both difficult targets for structural biology (size and oligomerization respectively). BRD4 was targeted with a well characterized inhibitor (JQ1), while VCP was investigated for binding with a known autophagy activator. Using HR-LiP we identify the binding site of the BRD4 inhibitor JQ1 in the full-length protein, which is typically too large to be used directly in with conventional methods. Further, we confirm binding of the autophagy activator to VCP at a site known to increase VCP activity. For both compounds tested, our data are in good accordance with orthogonal data obtained by other structural protein approaches including HDX-MS and “click” chemistry-based chemoproteomics. We demonstrate that HR-LiP can be used to dissect small molecule-protein binding events, including compound binding site prediction for protein targets classically considered to be difficult and time intensive. Further, we show that this data can be used to shed light on the biological mechanisms driving a phenotypic response. Given its biological power, broad applicability and ease of implementation, we envision the use of HR-LiP as a routine approach for target validation and lead optimization in small molecule drug discovery pipelines. Citation Format: Nigel Beaton, Roland Bruderer, Yuehan Feng, Jagat Adhikari, Ivan Cornella-Taracido, Lukas Reiter. High resolution limited proteolysis (HR-LiP), a novel structural proteomics approach for the prediction of small molecule-protein binding events [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 2024.

Docking cholesterol to integral membrane proteins with Rosetta.

Lipid molecules such as cholesterol interact with the surface of integral membrane proteins (IMP) in a mode different from drug-like molecules in a protein binding pocket. These differences are due to the lipid molecule's shape, the membrane's hydrophobic environment, and the lipid's orientation in the membrane. We can use the recent increase in experimental structures in complex with cholesterol to understand protein-cholesterol interactions. We developed the RosettaCholesterol protocol consisting of (1) a prediction phase using an energy grid to sample and score native-like binding poses and (2) a specificity filter to calculate the likelihood that a cholesterol interaction site may be specific. We used a multi-pronged benchmark (self-dock, flip-dock, cross-dock, and global-dock) of protein-cholesterol complexes to validate our method. RosettaCholesterol improved sampling and scoring of native poses over the standard RosettaLigand baseline method in 91% of cases and performs better regardless of benchmark complexity. On the β2AR, our method found one likely-specific site, which is described in the literature. The RosettaCholesterol protocol quantifies cholesterol binding site specificity. Our approach provides a starting point for high-throughput modeling and prediction of cholesterol binding sites for further experimental validation.

LINC01234 Sponging of the miR-513a-5p/AOX1 Axis is Upregulated in Osteoporosis and Regulates Osteogenic Differentiation of Bone Marrow Mesenchymal Stem Cells.

Non-coding RNAs, including long-chain non-coding RNA (lncRNA) and micro-RNA (miRNA), have been implicated in osteoporosis (OP) progression by regulating osteoblast-dependent bone metabolism. Herein, we investigated whether LINC01234, miR-513a-5p, and AOX1 regulate osteogenic differentiation and proliferation of human bone marrow mesenchymal stem cells (hMSCs). The expression of LINC01234, miR-513a-5p, and AOX1 was monitored using RT-qPCR or western blotting. Cell proliferation was assessed using a CCK8 assay. Alkaline phosphatase activity (ALP) and alizarin red dye staining were performed to determine osteogenic differentiation. Furthermore, the expression of osteoblast differentiation markers, such as ALP, BMP1 (bone morphogenetic protein 1), osteocalcin (OCN), and osteopontin (OPN), was determined by RT-qPCR. Luciferase reporter and RNA immunoprecipitation (RIP) assays were performed to verify the interplay between miR-513a-5p and LINC01234 or AOX1. Compared with the plasma of healthy controls, LINC01234 and AOX1 were highly expressed in the plasma of patients with OP, whereas miR-513a-5p showed low expression. In contrast, LINC01234 and AOX1 expression displayed a gradual decrease in induced differentiated hMSCs, while miR-513a-5p expression was upregulated with induction time. The predicted binding sites between miR-513a-5p and LINC01234 or AOX1 were verified by luciferase reporter and RIP assays. LINC01234 silencing induced osteogenic differentiation and proliferation in vitro, and miR-513a-5p silencing blunted osteogenic differentiation and proliferation modulated by LINC01234. AOX1 silencing caused by miR-513a-5p enhances osteogenic proliferation and differentiation. LINC01234 sponging of the miR-513a-5p/AOX1 axis impeded the osteogenic differentiation of hMSCs, favoring OP progression.

Endothelial KCa3.1 and KCa2.3 Mediate S1P (Sphingosine-1-Phosphate)-Dependent Vasodilation and Blood Pressure Homeostasis.

S1P (sphingosine-1-phosphate) has been reported to possess vasodilatory properties, but the underlying pathways are largely unknown. Isolated mouse mesenteric artery and endothelial cell models were used to determine S1P-induced vasodilation, intracellular calcium, membrane potentials, and calcium-activated potassium channels (KCa2.3 and KCa3.1 [endothelial small- and intermediate-conductance calcium-activated potassium channels]). Effect of deletion of endothelial S1PR1 (type 1 S1P receptor) on vasodilation and blood pressure was evaluated. Mesenteric arteries subjected to acute S1P stimulation displayed a dose-dependent vasodilation response, which was attenuated by blocking endothelial KCa2.3 or KCa3.1 channels. In cultured human umbilical vein endothelial cells, S1P stimulated immediate membrane potential hyperpolarization following activation of KCa2.3/KCa3.1 with elevated cytosolic Ca2+. Further, chronic S1P stimulation enhanced expression of KCa2.3 and KCa3.1 in human umbilical vein endothelial cells in dose- and time-dependent manners, which was abolished by disrupting either S1PR1-Ca2+ signaling or downstream Ca2+-activated calcineurin/NFAT (nuclear factor of activated T-cells) signaling. By combination of bioinformatics-based binding site prediction and chromatin immunoprecipitation assay, we revealed in human umbilical vein endothelial cells that chronic activation of S1P/S1PR1 promoted NFATc2 nuclear translocation and binding to promoter regions of KCa2.3 and KCa3.1 genes thus to upregulate transcription of these channels. Deletion of endothelial S1PR1 reduced expression of KCa2.3 and KCa3.1 in mesenteric arteries and exacerbated hypertension in mice with angiotensin II infusion. This study provides evidence for the mechanistic role of KCa2.3/KCa3.1-activated endothelium-dependent hyperpolarization in vasodilation and blood pressure homeostasis in response to S1P. This mechanistic demonstration would facilitate the development of new therapies for cardiovascular diseases associated with hypertension.

A Novel Deep Learning Method for Predicting RNA-Protein Binding Sites

The cell cycle and biological processes rely on RNA and RNA-binding protein (RBP) interactions. It is crucial to identify the binding sites on RNA. Various deep-learning methods have been used for RNA-binding site prediction. However, they cannot extract the hierarchical features of the RNA secondary structure. Therefore, this paper proposes HPNet, which can automatically identify RNA-binding sites and -binding preferences. HPNet performs feature learning from the two perspectives of the RNA sequence and the RNA secondary structure. A convolutional neural network (CNN), a deep-learning method, is used to learn RNA sequence features in HPNet. To capture the hierarchical information for RNA, we introduced DiffPool into HPNet, a differentiable pooling graph neural network (GNN). A CNN and DiffPool were combined to improve the binding site prediction accuracy by leveraging both RNA sequence features and hierarchical features of the RNA secondary structure. Binding preferences can be extracted based on model outputs and parameters. Overall, the experimental results showed that HPNet achieved a mean area under the curve (AUC) of 94.5% for the benchmark dataset, which was more accurate than the state-of-the-art methods. Moreover, these results demonstrate that the hierarchical features of RNA secondary structure play an essential role in selecting RNA-binding sites.

Genetic variants and haplotypes in fibulin-5 (FBLN5) are associated with pseudoexfoliation glaucoma but not with pseudoexfoliation syndrome.

Pseudoexfoliation (PEX) is a multifactorial age-related disease involving deposition of extracellular proteinaceous aggregates on anterior ocular tissues. The present study aims to identify functional variants in fibulin-5 (FBLN5) as risk factors for the development of PEX. Thirteen tag single-nucleotide polymorphisms (SNPs) in FBLN5 were genotyped using TaqMan SNP genotyping technology to identify association between SNPs of FBLN5 and PEX in an Indian cohort comprising 200 control and 273 PEX patients (169 PEXS and 104 PEXG). Functional analysis of risk variants was done through luciferase reporter assays and electrophoretic mobility shift assay (EMSA) using human lens epithelial cells. Genetic association and risk haplotype analysis showed a significant association of rs17732466:G>A (NC_000014.9:g.91913280G>A) and rs72705342:C>T (NC_000014.9:g.91890855C>T) within FBLN5 as risk factors with the advanced severe stage of the disease, pseudoexfoliation glaucoma (PEXG). Reporter assays showed allele-specific regulatory effect of rs72705342:C>T on gene expression, wherein, construct containing the risk allele showed a significant decrease in the reporter activity compared with the one with protective allele. EMSA further validated higher binding affinity of the risk variant to nuclear protein. In silico analysis predicted binding sites for two transcription factors, GR-α and TFII-I with risk allele at rs72705342:C>T, which were lost in the presence of protective allele. The EMSA showed probable binding of both these proteins to rs72705342. In conclusion, the present study identified the novel association of two genetic variants in FBLN5 with PEXG but not with PEXS, distinguishing between the early and the later forms of PEX. Further, rs72705342:C>T was found to be a functional variant.

MCNN: Multiple Convolutional Neural Networks for RNA-Protein Binding Sites Prediction.

Computational prediction of the RBP bound sites using features learned from existing annotation knowledge is an effective method because high-throughput experiments are complex, expensive and time-consuming. Many methods have been proposed to predict RNA-protein binding sites. However, the partial information of RNA sequence is not fully used. In this study, we propose multiple convolutional neural networks (MCNN) method, which predicts RNA-protein binding sites by integrating multiple convolutional neural networks constructed by RNA sequence information extracted from windows with different lengths. First, MCNN trains multiple CNNs base on RNA sequences extracted by different window lengths. Second, MCNN can extract more binding patterns of RBPs by combining these trained multiple CNNs previously. Third, MCNN only uses RNA base sequence information for RNA-protein binding sites prediction, which extracts sequence binding features and predicts the result with same architecture. This avoids the information loss of feature extraction step. Our proposed MCNN demonstrates a competitive performance comparing with other methods on a large-scale dataset derived from CLIP-seq, which is an effective method for RNA-protein binding sites prediction. The source code of our proposed MCNN method can be found in https://github.com/biomg/MCNN.

Using Attribution Sequence Alignment to Interpret Deep Learning Models for miRNA Binding Site Prediction

MicroRNAs (miRNAs) are small non-coding RNAs that play a central role in the post-transcriptional regulation of biological processes. miRNAs regulate transcripts through direct binding involving the Argonaute protein family. The exact rules of binding are not known, and several in silico miRNA target prediction methods have been developed to date. Deep learning has recently revolutionized miRNA target prediction. However, the higher predictive power comes with a decreased ability to interpret increasingly complex models. Here, we present a novel interpretation technique, called attribution sequence alignment, for miRNA target site prediction models that can interpret such deep learning models on a two-dimensional representation of miRNA and putative target sequence. Our method produces a human readable visual representation of miRNA:target interactions and can be used as a proxy for the further interpretation of biological concepts learned by the neural network. We demonstrate applications of this method in the clustering of experimental data into binding classes, as well as using the method to narrow down predicted miRNA binding sites on long transcript sequences. Importantly, the presented method works with any neural network model trained on a two-dimensional representation of interactions and can be easily extended to further domains such as protein-protein interactions.