A systematic method for solving data imbalance in CRISPR off-target prediction tasks.

Accurately identifying potential off-target sites in the CRISPR/Cas9 system is crucial for improving the efficiency and safety of editing. However, the imbalance of available off-target datasets has posed a major obstacle in enhancing prediction performance. Despite several prediction models have been developed to address this issue, there remains a lack of systematic research on handling data imbalance in off-target prediction. This article systematically investigates the data imbalance issue in off-target datasets and explores numerous methods to process data imbalance from a novel perspective. First, we highlight the impact of the imbalance problem on off-target prediction tasks by determining the imbalance ratios present in these datasets. Then, we provide a comprehensive review of various sampling techniques and cost-sensitive methods to mitigate class imbalance in off-target datasets. Finally, systematic experiments are conducted on several state-of-the-art prediction models to illustrate the impact of applying data imbalance solutions. The results show that class imbalance processing methods significantly improve the off-target prediction capabilities of the models across multiple testing datasets. The code and datasets used in this study are available at https://github.com/gzrgzx/CRISPR_Data_Imbalance.

Global Atlas of Methane Metabolism Marker Genes in Soil.

Methane, a greenhouse gas, plays a pivotal role in the global carbon cycle, influencing the Earth's climate. Only a limited number of microorganisms control the flux of biologically produced methane in nature, including methane-oxidizing bacteria, anaerobic methanotrophic archaea, and methanogenic archaea. Although previous studies have revealed the spatial and temporal distribution characteristics of methane-metabolizing microorganisms in local regions by using the marker genes pmoA or mcrA, their biogeographical patterns and environmental drivers remain largely unknown at a global scale. Here, we used 3419 metagenomes generated from georeferenced soil samples to examine the global patterns of methane metabolism marker gene abundances in soil, which generally represent the global distribution of methane-metabolizing microorganisms. The resulting maps revealed notable latitudinal trends in the abundances of methane-metabolizing microorganisms across global soils, with higher abundances in the sub-Arctic, sub-Antarctic, and tropical rainforest regions than in temperate regions. The variations in global abundances of methane-metabolizing microorganisms were primarily governed by vegetation cover. Our high-resolution global maps of methane-metabolizing microorganisms will provide valuable information for the prediction of biogenic methane emissions under current and future climate scenarios.

Transformer-based anti-noise models for CRISPR-Cas9 off-target activities prediction.

The off-target effect occurring in the CRISPR-Cas9 system has been a challenging problem for the practical application of this gene editing technology. In recent years, various prediction models have been proposed to predict potential off-target activities. However, most of the existing prediction methods do not fully exploit guide RNA (gRNA) and DNA sequence pair information effectively. In addition, available prediction methods usually ignore the noise effect in original off-target datasets. To address these issues, we design a novel coding scheme, which considers the key features of mismatch type, mismatch location and the gRNA-DNA sequence pair information. Furthermore, a transformer-based anti-noise model called CrisprDNT is developed to solve the noise problem that exists in the off-target data. Experimental results of eight existing datasets demonstrate that the method with the inclusion of the anti-noise loss functions is superior to available state-of-the-art prediction methods. CrisprDNT is available at https://github.com/gzrgzx/CrisprDNT.

TransCrispr: Transformer Based Hybrid Model for Predicting CRISPR/Cas9 Single Guide RNA Cleavage Efficiency.

CRISPR/Cas9 is a widely used genome editing tool for site-directed modification of deoxyribonucleic acid (DNA) nucleotide sequences. However, how to accurately predict and evaluate the on- and off-target effects of single guide RNA (sgRNA) is one of the key problems for CRISPR/Cas9 system. Using computational methods to obtain high cell-specific sensitivity and specificity is a prerequisite for the optimal design of sgRNAs. Inspired by the work of predecessors, we found that sgRNA on-target knockout efficacy was not only related to the original sequence but also affected by important biological features. Hence, we introduce a novel approach called TransCrispr, which integrates Transformer and convolutional neural network (CNN) architecture to predict sgRNA knockout efficacy. Firstly, we encode the sequence data and send the transformed sgRNA sequence, positional information, and biological features into the network as input. Then, the convolutional neural network will automatically learn an appropriate feature representation for the sgRNA sequence and combine it with the positional information for self-attention learning of the Transformer. Finally, a regression score is generated by predicting biological features. Experiments on seven public datasets illustrate that TransCrispr outperforms state-of-the-art methods in terms of prediction accuracy and generalization ability.

GraphDPA: Predicting drug-pathway associations by graph convolutional networks.

Pathway-based drug discovery is a promising strategy for the discovery of drugs with low toxicity and side effects. However, identifying the associations between drug and targeting pathways is challenging for this method. The formation of various biomolecular interaction databases and the development of neural network technology provide new ways for the large-scale prediction of drug-pathway associations. This article proposes a new model called GraphDPA, which represents the drug and pathway-gene association as a graph. We use graph convolutional networks (GCN) to learn the features of the drug and pathway and predict the drug-pathway association. The results show that GraphDPA can predict drug-pathway associations with high accuracy, which verify the potential of the GCN in drug discovery.

Effective use of sequence information to predict CRISPR-Cas9 off-target.

The CRISPR/Cas9 gene-editing system is the third-generation gene-editing technology that has been widely used in biomedical applications. However, off-target effects occurring CRISPR/Cas9 system has been a challenging problem it faces in practical applications. Although many predictive models have been developed to predict off-target activities, current models do not effectively use sequence pair information. There is still room for improved accuracy. This study aims to effectively use sequence pair information to improve the model's performance for predicting off-target activities. We propose a new coding scheme for coding sequence pairs and design a new model called CRISPR-IP for predicting off-target activity. Our coding scheme distinguishes regions with different functions in the sequence pairs through the function channel. Moreover, it distinguishes between bases and base pairs using type channels, effectively representing the sequence pair information. The CRISPR-IP model is based on CNN, BiLSTM, and the attention layer to learn features of sequence pairs. We performed performance verification on two data sets and found that our coding scheme can represent sequence pair information effectively, and the CRISPR-IP model performance is better than others. Data and source codes are available at https://github.com/BioinfoVirgo/CRISPR-IP.

AttCRISPR: a spacetime interpretable model for prediction of sgRNA on-target activity.

BACKGROUND: More and more Cas9 variants with higher specificity are developed to avoid the off-target effect, which brings a significant volume of experimental data. Conventional machine learning performs poorly on these datasets, while the methods based on deep learning often lack interpretability, which makes researchers have to trade-off accuracy and interpretability. It is necessary to develop a method that can not only match deep learning-based methods in performance but also with good interpretability that can be comparable to conventional machine learning methods. RESULTS: To overcome these problems, we propose an intrinsically interpretable method called AttCRISPR based on deep learning to predict the on-target activity. The advantage of AttCRISPR lies in using the ensemble learning strategy to stack available encoding-based methods and embedding-based methods with strong interpretability. Comparison with the state-of-the-art methods using WT-SpCas9, eSpCas9(1.1), SpCas9-HF1 datasets, AttCRISPR can achieve an average Spearman value of 0.872, 0.867, 0.867, respectively on several public datasets, which is superior to these methods. Furthermore, benefits from two attention modules-one spatial and one temporal, AttCRISPR has good interpretability. Through these modules, we can understand the decisions made by AttCRISPR at both global and local levels without other post hoc explanations techniques. CONCLUSION: With the trained models, we reveal the preference for each position-dependent nucleotide on the sgRNA (short guide RNA) sequence in each dataset at a global level. And at a local level, we prove that the interpretability of AttCRISPR can be used to guide the researchers to design sgRNA with higher activity.

ISFMDA: Learning Interactions of Selected Features-Based Method for Predicting Potential MicroRNA-Disease Associations.

Prediction of potential microRNA-disease associations is one of the important tasks in computational biology fields. Mining more sophisticated features can improve the performance of the prediction methods. This article proposes a novel algorithm (ISFMDA) that can effectively learn low- or high-order interactions of recursive feature elimination selected features by an extreme gradient boosting, a factorization machine, and a deep neural network. As a result, ISFMDA can obtain an area under receiver operating characteristic curve (AUROC) of 0.9342 ± 0.0007 in fivefold cross-validation tests with 51.25% of original features, which verifies the effectiveness of the methods.

ANMDA: anti-noise based computational model for predicting potential miRNA-disease associations.

BACKGROUND: A growing proportion of research has proved that microRNAs (miRNAs) can regulate the function of target genes and have close relations with various diseases. Developing computational methods to exploit more potential miRNA-disease associations can provide clues for further functional research. RESULTS: Inspired by the work of predecessors, we discover that the noise hiding in the data can affect the prediction performance and then propose an anti-noise algorithm (ANMDA) to predict potential miRNA-disease associations. Firstly, we calculate the similarity in miRNAs and diseases to construct features and obtain positive samples according to the Human MicroRNA Disease Database version 2.0 (HMDD v2.0). Then, we apply k-means on the undetected miRNA-disease associations and sample the negative examples equally from the k-cluster. Further, we construct several data subsets through sampling with replacement to feed on the light gradient boosting machine (LightGBM) method. Finally, the voting method is applied to predict potential miRNA-disease relationships. As a result, ANMDA can achieve an area under the receiver operating characteristic curve (AUROC) of 0.9373 ± 0.0005 in five-fold cross-validation, which is superior to several published methods. In addition, we analyze the predicted miRNA-disease associations with high probability and compare them with the data in HMDD v3.0 in the case study. The results show ANMDA is a novel and practical algorithm that can be used to infer potential miRNA-disease associations. CONCLUSION: The results indicate the noise hiding in the data has an obvious impact on predicting potential miRNA-disease associations. We believe ANMDA can achieve better results from this task with more methods used in dealing with the data noise.

HColonDB: A Database for Human Colon Cancer Research.

HColonDB (Human Colon cancer Database) is an important database which integrates genes, pathways, networks, drugs, and other information related to colon cancer. The purpose of the database is to provide a platform for the systematic research of colon cancer. The relationships between genes and pathways, genes and networks, and networks and pathways are obtained from the database KEGG. Furthermore, the information of the drugs used to treat colon cancer is available in HColonDB, which is collected and organized from DrugBank and PubChem database. In brief, we have summarized 81 genes, 112 pathways, 108 networks, and 15 drugs associated with colon cancer. The current version of HColonDB contains 322 associations between genes and pathways, 242 associations between genes and networks, and 68 associations between networks and pathways. In addition, HColonDB provides a friendly interface for users to browse and search. We hope that the database can make it more convenient for researchers to get the data they need and help in the treatment of colon cancer.

Plant natural product plumbagin presents potent inhibitory effect on human cytochrome P450 2J2 enzyme.

BACKGROUND: Cytochrome P450 2J2 (CYP2J2) is not only highly expressed in many kinds of human tumors, but also promotes tumor cell growth via regulating the metabolism of arachidonic acids. CYP2J2 inhibitors can significantly reduce proliferation, migration and promote apoptosis of tumor cells by inhibiting epoxyeicosatrienoic acids (EETs) biosynthesis. Therefore screening CYP2J2 inhibitors is a significant way for the development of anti-cancer drug. PURPOSE: The aim of this study was to identify a new CYP2J2 inhibitor from fifty natural compounds obtained from plants. STUDY DESIGN: CYP2J2 inhibitor was screened from a natural compounds library and further the inhibitory manner and mechanism were evaluated. Its cytotoxicity against HepG2 and SMMC-7721 cell lines was also estimated. METHODS: The inhibitory effect was evaluated in rat liver microsomes (RLMs), human liver microsomes (HLMs) and recombinant CYP2J2 (rCYP2J2), using astemizole as a probe substrate and inhibitory mechanism was illustrated through molecular docking. The cytotoxicity was detected using SRB. RESULTS: In all candidates, plumbagin showed the strongest inhibitory effect on the CYP2J2-mediated astemizole O-demethylation activity. Further study revealed that plumbagin potently inhibited CYP2J2 activity with IC50 value at 3.82 µM, 3.37 µM and 1.17 µM in RLMs, HLMs and rCYP2J2, respectively. Enzyme kinetic studies showed that plumbagin was a mixed-type inhibitor of CYP2J2 in HLMs and rCYP2J2 with Ki value of 1.88 µM and 0.92 µM, respectively. Docking data presented that plumbagin interacted with CYP2J2 mainly through GLU 222 and ALA 223. Moreover, plumbagin showed strongly cytotoxic effects on hepatoma cell lines, such as HepG2 and SMMC-7721, with lower toxicity on rat primary hepatocytes. Plumbagin had no effect on the protein expression of CYP2J2 in HepG2 and SMMC-7721, while down-regulated the mRNA level of anti-apoptosis protein Bcl-2. CONCLUSION: This study found out a new CYP2J2 inhibitor plumbagin from fifty natural compounds. Plumbagin presented a potential of anti-cancer pharmacological activity.

Prediction of Central Nervous System Side Effects Through Drug Permeability to Blood-Brain Barrier and Recommendation Algorithm.

Drug side effects are one of the public health concerns. Using powerful machine-learning methods to predict potential side effects before the drugs reach the clinical stages is of great importance to reduce time consumption and protect the security of patients. Recently, researchers have proved that the central nervous system (CNS) side effects of a drug are closely related to its permeability to the blood-brain barrier (BBB). Inspired by this, we proposed an extended neighborhood-based recommendation method to predict CNS side effects using drug permeability to the BBB and other known features of drug. To the best of our knowledge, this is the first attempt to predict CNS side effects considering drug permeability to the BBB. Computational experiments demonstrated that drug permeability to the BBB is an important factor in CNS side effects prediction. Moreover, we built an ensemble recommendation model and obtained higher AUC score (area under the receiver operating characteristic curve) and AUPR score (area under the precision-recall curve) on the data set of CNS side effects by integrating various features of drug.

Prediction of drug-pathway interaction pairs with a disease-combined LSA-PU-KNN method.

Prediction of new associations between drugs and targeting pathways can provide valuable clues for drug discovery & development. However, information integration and a class-imbalance problem are important challenges for available prediction methods. This paper proposes a prediction of potential associations between drugs and pathways based on a disease-related LSA-PU-KNN method. Firstly, we built a drug-disease-pathway network and combined the drug-disease and pathway-disease features obtained by different types of feature profiles. Then we applied a latent semantic analysis (LSA) method to perform dimension reduction by combining positive-unlabeled (PU) learning and k nearest neighbors (KNN) method. The experimental results showed that our method can achieve a higher AUC (the area under the ROC curve) and AUPR (the area under the PR curve) than other typical methods. Furthermore, some interesting drug-pathway interaction pairs were identified and validated.

Exploration and analysis of drug modes of action through feature integration.

Identifying drug modes of action (MoA) is of paramount importance for having a good grasp of drug indications in clinical tests. Anticipating MoA can help to discover new uses for approved drugs. Here we first used a drug-set enrichment analysis method to discover significant biological activities in every mode of action category. Then, we proposed a new computational model, a probability ensemble approach based on Bayesian network theory, which integrated chemical, therapeutic, genomic and phenotypic properties of over a thousand of FDA approved drugs to assist with the prediction of MoA. 10-fold cross validation tests demonstrate that this method can achieve better performances than four other methods with the area under the receiver operating characteristic (ROC) curves. Finally, we further conducted a large-scale prediction for drug-MoA pairs. Using the Cardiovascular Agents category as an example, several predicted drug-MoA pairs were supported by literature resources.

Identification of Genes Involved in Breast Cancer Metastasis by Integrating Protein-Protein Interaction Information with Expression Data.

The selection of relevant genes for breast cancer metastasis is critical for the treatment and prognosis of cancer patients. Although much effort has been devoted to the gene selection procedures by use of different statistical analysis methods or computational techniques, the interpretation of the variables in the resulting survival models has been limited so far. This article proposes a new Random Forest (RF)-based algorithm to identify important variables highly related with breast cancer metastasis, which is based on the important scores of two variable selection algorithms, including the mean decrease Gini (MDG) criteria of Random Forest and the GeneRank algorithm with protein-protein interaction (PPI) information. The new gene selection algorithm can be called PPIRF. The improved prediction accuracy fully illustrated the reliability and high interpretability of gene list selected by the PPIRF approach.

Discovery and Optimization of N-Substituted 2-(4-pyridinyl)thiazole carboxamides against Tumor Growth through Regulating Angiogenesis Signaling Pathways.

Inhibition of angiogenesis is considered as one of the desirable pathways for the treatment of tumor growth and metastasis. Herein we demonstrated that a series of pyridinyl-thiazolyl carboxamide derivatives were designed, synthesized and examined against angiogenesis through a colony formation and migration assays of human umbilical vein endothelial cells (HUVECs) in vitro. A structure-activity relationship (SAR) study was carried out and optimization toward this series of compounds resulted in the discovery of N-(3-methoxyphenyl)-4-methyl-2-(2-propyl-4-pyridinyl)thiazole-5-carboxamide (3k). The results indicated that compound 3k showed similar or better effects compared to Vandetanib in suppressing HUVECs colony formation and migration as well as VEGF-induced angiogenesis in the aortic ring spreading model and chick embryo chorioallantoic membrane (CAM) model. More importantly, compound 3k also strongly blocked tumor growth with the dosage of 30 mg/kg/day, and subsequent mechanism exploration suggested that this series of compounds took effect mainly through angiogenesis signaling pathways. Together, these results suggested compound 3k may serve as a lead for a novel class of angiogenesis inhibitors for cancer treatments.

DSEP: A Tool Implementing Novel Method to Predict Side Effects of Drugs.

Drug side effects, or adverse drug reactions, have become a focus of public health concern. Anticipating side effects before the drugs are granted marketing authorization for clinical use can help reduce health threats. An increasing need for methods and tools that facilitate side-effect prediction still remains. Here, we present DSEP, which is a tool that is able to analyze chemistry files to predict side effects of drugs that are under development and have not been included into any database. Meanwhile, DSEP provides three computational methods, one of which is a novel method proposed by us. The method can obtain higher AUC(0.8927) and AUPR(0.4143) scores than previous work. The advantage characteristic and method made DSEP a useful tool to predict potential side effects for a given drug or compound. We use DSEP to conduct uncharacterized drugs' side-effect prediction and confirm interesting results.

Prediction of drug's Anatomical Therapeutic Chemical (ATC) code by integrating drug-domain network.

Predicting Anatomical Therapeutic Chemical (ATC) code of drugs is of vital importance for drug classification and repositioning. Discovering new association information related to drugs and ATC codes is still difficult for this topic. We propose a novel method named drug-domain hybrid (dD-Hybrid) incorporating drug-domain interaction network information into prediction models to predict drug's ATC codes. It is based on the assumption that drugs interacting with the same domain tend to share therapeutic effects. The results demonstrated dD-Hybrid has comparable performance to other methods on the gold standard dataset. Further, several new predicted drug-ATC pairs have been verified by experiments, which offer a novel way to utilize drugs for new purposes effectively.

Inferring Association between Compound and Pathway with an Improved Ensemble Learning Method.

Emergence of compound molecular data coupled to pathway information offers the possibility of using machine learning methods for compound-pathway associations' inference. To provide insights into the global relationship between compounds and their affected pathways, a improved Rotation Forest ensemble learning method called RGRF (Relief & GBSSL - Rotation Forest) was proposed to predict their potential associations. The main characteristic of the RGRF lies in using the Relief algorithm for feature extraction and regarding the Graph-Based Semi-Supervised Learning method as classifier. By incorporating the chemical structure information, drug mode of action information and genomic space information, our method can achieve a better precision and flexibility on compound-pathway prediction. Moreover, several new compound-pathway associations that having the potential for further clinical investigation have been identified by database searching. In the end, a prediction tool was developed using RGRF algorithm, which can predict the interactions between pathways and all of the compounds in cMap database.

Inferring novel indications of approved drugs via a learning method with local and global consistency.

Inferring new indication of approved drugs is critical not only for the elucidation of the interaction mechanisms between these drugs and their associated diseases, but also for the development of drug therapy for various human diseases. This paper proposes a network-based approach to reveal the association between 52 human diseases and potential therapeutic drugs based on multiple types of data. The advantage of the approach is that it can obtain the global relevance features for each drug-disease pair in the network by the learning local and global consistency method (LLGC). Cross-validation tests results demonstrate the proposed approach can achieve better performance comparing with previous methods. More importantly, it provides a promising strategy to maximize the value of therapeutic drugs and offer safe and effective treatments for different diseases.