Background: CRISPR/Cas9 system, as the third-generation genome editing technology, has been widely applied in target gene repair and gene expression regulation. Selection of appropriate sgRNA can improve the on-target knockout efficacy of CRISPR/Cas9 system with high sensitivity and specificity. However, when CRISPR/Cas9 system is operating, unexpected cleavage may occur at some sites, known as off-target. Presently, a number of prediction methods have been developed to predict the off-target propensity of sgRNA at specific DNA fragments. Most of them use artificial feature extraction operations and machine learning techniques to obtain off-target scores. With the rapid expansion of off-target data and the rapid development of deep learning theory, the existing prediction methods can no longer satisfy the prediction accuracy at the clinical level. Results: Here, we propose a prediction method named CnnCrispr to predict the off-target propensity of sgRNA at specific DNA fragments. CnnCrispr automatically trains the sequence features of sgRNA-DNA pairs with GloVe model, and embeds the trained word vector matrix into the deep learning model including biLSTM and CNN with five hidden layers. We conducted performance verification on the data set provided by DeepCrispr, and found that the auROC and auPRC in the “leave-one-sgRNA-out” cross validation could reach 0.957 and 0.429 respectively (the Pearson value and spearman value could reach 0.495 and 0.151 respectively under the same settings). Conclusion: Our results show that CnnCrispr has better classification and regression performance than the existing states-of-art models. The code for CnnCrispr can be freely downloaded from https://github.com/LQYoLH/CnnCrispr.
With the rapid development of high-throughput sequencing technology, a large number of transcript sequences have been discovered, and how to identify long non-coding RNAs (lncRNAs) from transcripts is a challenging task. The identification and inclusion of lncRNAs not only can more clearly help us to understand life activities themselves, but can also help humans further explore and study the disease at the molecular level. At present, the detection of lncRNAs mainly includes two forms of calculation and experiment. Due to the limitations of bio sequencing technology and ineluctable errors in sequencing processes, the detection effect of these methods is not very satisfactory. In this paper, we constructed a deep-learning model to effectively distinguish lncRNAs from mRNAs. We used k-mer embedding vectors obtained through training the GloVe algorithm as input features and set up the deep learning framework to include a bidirectional long short-term memory model (BLSTM) layer and a convolutional neural network (CNN) layer with three additional hidden layers. By testing our model, we have found that it obtained the best values of 97.9%, 96.4% and 99.0% in F1score, accuracy and auROC, respectively, which showed better classification performance than the traditional PLEK, CNCI and CPC methods for identifying lncRNAs. We hope that our model will provide effective help in distinguishing mature mRNAs from lncRNAs, and become a potential tool to help humans understand and detect the diseases associated with lncRNAs.
Piwi-interacting RNA (piRNA) is a class of small non-coding RNAs about 24 to 32 nucleotides long, associated with PIWI proteins, which are involved in germline development, transposon silencing, and epigenetic regulation. Identification of piRNA loci on the genome is very useful for further studies in the biogenesis and function of piRNAs. To accomplish this, we applied the computational biology tool Teiresias to identify motifs of variable length appearing frequently in mouse piRNA and non-piRNA sequences, respectively, and then proposed an algorithm for piRNA identification based on motif discovery, termed “Pibomd” by using these sequence motifs as features in the Support Vector Machine (SVM) algorithm, a sensitivity of 91.48% and a specificity of 89.76% on a mouse test dataset could be achieved, much better results than those reported in previously published algorithms. We also trained an unbalanced SVM classifier (named as “Asym-Pibomd”) that provided a higher specificity (96.2%) and a lower sensitivity (72.68%) than Pibomd. Inspite of the predicted ACC being less than that of Pibomd, the predicted ACC (84.44%) of Asym-Pibomd is about ten percent more than that obtained using the k-mer method. Further analysis of the motif positions on the piRNA sequences showed that the piRNA sequences may contain information at the 5′- and/or 3′-end recognized by the piRNA processing apparatus of actual piRNA precursors. Furthermore, this prediction method can be found on a user-friendly web server found at http://app.aporc.org/Pibomd/.
Clustered regularly interspaced short palindromic repeats (CRISPR) gene-editing technology has transformed molecular biology. Predicting editing efficiency is crucial for optimization, and numerous computational models have been created. However, many current models struggle to generalize across diverse editing systems, often experiencing performance drops with varying conditions or systems. Additionally, most models focus on ribonucleic acid (RNA) sequence and thermodynamic features, overlooking the importance of secondary structure information. Here, we present the first graph-based model (Graph-CRISPR) that integrates both sequence and secondary structure features of single guide RNA enhancing editing efficiency prediction. Tests show Graph-CRISPR consistently surpasses baseline models across systems like CRISPR-Cas9, prime editing, and base editing. It also demonstrates strong resilience, maintaining robust performance under varying experimental conditions. This work highlights the potential of integrating sequence and structural information through graph-based modeling to enhance predictive accuracy and adaptability in gene editing applications. The datasets and source codes are publicly available at: https://github.com/MoonLBH/Graph-CRISPR
Tissues are constituted of heterogeneous cell types. Although single-cell RNA sequencing has paved the way to a deeper understanding of organismal cellular composition, the high cost and technical noise have prevented its wide application. As an alternative, computational deconvolution of bulk tissues can be a cost-effective solution. In this study, we propose DecOT, a deconvolution method that uses the Wasserstein distance as a loss and applies scRNA-seq data as references to characterize the cell type composition from bulk tissue RNA-seq data. The Wasserstein loss in DecOT is able to utilize additional information from gene space. DecOT also applies an ensemble framework to integrate deconvolution results from multiple individuals’ references to mitigate the individual/batch effect. By benchmarking DecOT with four recently proposed square loss-based methods on pseudo-bulk data from four different single-cell data sets and real pancreatic islet bulk samples, we show that DecOT outperforms other methods and the ensemble framework is robust to the choice of references.
Background CRISPR/Cas9 system, as the third-generation genome editing technology, has been widely applied in target gene repair and gene expression regulation. Selection of appropriate sgRNA can improve the on-target knockout efficacy of CRISPR/Cas9 system with high sensitivity and specificity. However, when CRISPR/Cas9 system is operating, unexpected cleavage may occur at some sites, known as off-target. Presently, a number of prediction methods have been developed to predict the off-target propensity of sgRNA at specific DNA fragment. Most of them use artificial feature extraction operations and machine learning techniques to obtain off-target scores. With the rapid expansion of off-target data and the rapid development of deep learning theory, the existing prediction methods can no longer satisfy the prediction accuracy at the clinical level. Results Here, we propose a prediction method named CnnCrispr to predict the off-target propensity of sgRNA at specific DNA fragments. CnnCrispr automatically trains the sequence features of sgRNA-DNA pairs with GloVe model, and embeds the trained word vector matrix into the deep learning model including biLSTM and CNN with five hidden layers. We conducted performance verification on the data set provided by DeepCrispr, and found that the auROC and auPRC in the "leave-one-sgRNA-out" cross validation could reach 0.957 and 0.429 respectively (the pearson value and spearman value could reach 0.495 and 0.151 respectively under the same settings). Conclusion Our results show that CnnCrispr has better classification and regression performance than the existing states-of-art models. The code of CnnCrispr can be freely downloaded from https://github.com/LQYoLH/CnnCrispr.