Using data from gene combination double knockout (CDKO) experiments, top ranked synthetic lethal (SL) gene pairs were highly inconsistent among different SL scores. This leads to a significant concern that SL prediction models highly depend on SL scores. In this paper, we introduce a new gene combination effect (GCE) measurement, log-fold change of dual-gRNA expression before and after CRISPR-cas9 lentivirus transfection. We show it is a direct and highly consistent measurement of GCE in all CDKO experiments. We therefore develop a multi-layer encoder model for individual sample specific GCE prediction, MLEC-iGeneCombo. Under a deep learning framework, MLEC-iGeneCombo is a systems biology model that contains sample specific multi-omics encoder, network encoder and cell-line encoder. For the first time, MLEC-iGeneCombo predicts GCE for a new cell. Using data from 18 CDKO experiments, MLEC-iGeneCombo achieves an average GCE prediction performance, 71.9%. All three encoders significantly improve the model's prediction performance (p[Formula: see text]), and their combined use yields the best GCE prediction performance. Our source code is available at https://github.com/karenyun/MLEC-iGeneCombo.

Synthetic lethality (SL) has shown great promise for the discovery of novel targets in cancer. CRISPR double-knockout (CDKO) technologies can only screen several hundred genes and their combinations, but not genome-wide. Therefore, good SL prediction models are highly needed for genes and gene pairs selection in CDKO experiments. However, lack of scalable SL properties prevents generalizability of SL interactions to out-of-sample data, thereby hindering modeling efforts. In this paper, we recognize that SL connectivity is a scalable and generalizable SL property. We develop a novel two-step multilayer encoder for individual sample-specific SL prediction model (MLEC-iSL), which predicts SL connectivity first and SL interactions subsequently. MLEC-iSL has three encoders, namely, gene, graph, and transformer encoders. MLEC-iSL achieves high SL prediction performance in K562 (AUPR, 0.73; AUC, 0.72) and Jurkat (AUPR, 0.73; AUC, 0.71) cells, while no existing methods exceed 0.62 AUPR and AUC. The prediction performance of MLEC-iSL is validated in a CDKO experiment in 22Rv1 cells, yielding a 46.8% SL rate among 987 selected gene pairs. The screen also reveals SL dependency between apoptosis and mitosis cell death pathways.

MOTIVATION: Clustered regularly interspaced short palindromic repeats (CRISPR)-based genetic perturbation screen is a powerful tool to probe gene function. However, experimental noises, especially for the lowly expressed genes, need to be accounted for to maintain proper control of false positive rate. METHODS: We develop a statistical method, named CRISPR screen with Expression Data Analysis (CEDA), to integrate gene expression profiles and CRISPR screen data for identifying essential genes. CEDA stratifies genes based on expression level and adopts a three-component mixture model for the log-fold change of single-guide RNAs (sgRNAs). Empirical Bayesian prior and expectation-maximization algorithm are used for parameter estimation and false discovery rate inference. RESULTS: Taking advantage of gene expression data, CEDA identifies essential genes with higher expression. Compared to existing methods, CEDA shows comparable reliability but higher sensitivity in detecting essential genes with moderate sgRNA fold change. Therefore, using the same CRISPR data, CEDA generates an additional hit gene list. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Abstract Image captioning means generate descriptive sentences from a query image automatically. It has recently received widespread attention from the computer vision and natural language processing communities as an emerging visual task. Currently, both components have evolved considerably by exploiting object regions, attributes, attention mechanism methods, entity recognition with novelties, and training strategies. However, despite the impressive results, the research has not yet come to a conclusive answer. This survey aims to provide a comprehensive overview of image captioning methods, from technical architectures to benchmark datasets, evaluation metrics, and comparison of state‐of‐the‐art methods. In particular, image captioning methods are divided into different categories based on the technique adopted. Representative methods in each class are summarized, and their advantages and limitations are discussed. Moreover, many related state‐of‐the‐art studies were quantitatively compared to determine the recent trends and future directions in image captioning. The ultimate goal of this work is to serve as a tool for understanding the existing literature and highlighting future directions in the area of image captioning for Computer Vision and Natural Language Processing communities may benefit from.

Single-cell mass cytometry, also known as cytometry by time of flight (CyTOF) is a powerful high-throughput technology that allows analysis of up to 50 protein markers per cell for the quantification and classification of single cells. Traditional manual gating utilized to identify new cell populations has been inadequate, inefficient, unreliable, and difficult to use, and no algorithms to identify both calibration and new cell populations has been well established. A deep learning with graphic cluster (DGCyTOF) visualization is developed as a new integrated embedding visualization approach in identifying canonical and new cell types. The DGCyTOF combines deep-learning classification and hierarchical stable-clustering methods to sequentially build a tri-layer construct for known cell types and the identification of new cell types. First, deep classification learning is constructed to distinguish calibration cell populations from all cells by softmax classification assignment under a probability threshold, and graph embedding clustering is then used to identify new cell populations sequentially. In the middle of two-layer, cell labels are automatically adjusted between new and unknown cell populations via a feedback loop using an iteration calibration system to reduce the rate of error in the identification of cell types, and a 3-dimensional (3D) visualization platform is finally developed to display the cell clusters with all cell-population types annotated. Utilizing two benchmark CyTOF databases comprising up to 43 million cells, we compared accuracy and speed in the identification of cell types among DGCyTOF, DeepCyTOF, and other technologies including dimension reduction with clustering, including Principal Component Analysis (PCA), Factor Analysis (FA), Independent Component Analysis (ICA), Isometric Feature Mapping (Isomap), t-distributed Stochastic Neighbor Embedding (t-SNE), and Uniform Manifold Approximation and Projection (UMAP) with k-means clustering and Gaussian mixture clustering. We observed the DGCyTOF represents a robust complete learning system with high accuracy, speed and visualization by eight measurement criteria. The DGCyTOF displayed F-scores of 0.9921 for CyTOF1 and 0.9992 for CyTOF2 datasets, whereas those scores were only 0.507 and 0.529 for the t-SNE+k-means; 0.565 and 0.59, for UMAP+ k-means. Comparison of DGCyTOF with t-SNE and UMAP visualization in accuracy demonstrated its approximately 35% superiority in predicting cell types. In addition, observation of cell-population distribution was more intuitive in the 3D visualization in DGCyTOF than t-SNE and UMAP visualization. The DGCyTOF model can automatically assign known labels to single cells with high accuracy using deep-learning classification assembling with traditional graph-clustering and dimension-reduction strategies. Guided by a calibration system, the model seeks optimal accuracy balance among calibration cell populations and unknown cell types, yielding a complete and robust learning system that is highly accurate in the identification of cell populations compared to results using other methods in the analysis of single-cell CyTOF data. Application of the DGCyTOF method to identify cell populations could be extended to the analysis of single-cell RNASeq data and other omics data.

Cancer is a complex disease with usually multiple disease mechanisms. Target combination is a better strategy than a single target in developing cancer therapies. However, target combinations are generally more difficult to be predicted. Current CRISPR-cas9 technology enables genome-wide screening for potential targets, but only a handful of genes have been screend as target combinations. Thus, an effective computational approach for selecting candidate target combinations is highly desirable. Selected target combinations also need to be translational between cell lines and cancer patients. We have therefore developed DSCN (double-target selection guided by CRISPR screening and network), a method that matches expression levels in patients and gene essentialities in cell lines through spectral-clustered protein-protein interaction (PPI) network. In DSCN, a sub-sampling approach is developed to model first-target knockdown and its impact on the PPI network, and it also facilitates the selection of a second target. Our analysis first demonstrated a high correlation of the DSCN sub-sampling-based gene knockdown model and its predicted differential gene expressions using observed gene expression in 22 pancreatic cell lines before and after MAP2K1 and MAP2K2 inhibition (R2 = 0.75). In DSCN algorithm, various scoring schemes were evaluated. The 'diffusion-path' method showed the most significant statistical power of differentialting known synthetic lethal (SL) versus non-SL gene pairs (P = 0.001) in pancreatic cancer. The superior performance of DSCN over existing network-based algorithms, such as OptiCon and VIPER, in the selection of target combinations is attributable to its ability to calculate combinations for any gene pairs, whereas other approaches focus on the combinations among optimized regulators in the network. DSCN's computational speed is also at least ten times fast than that of other methods. Finally, in applying DSCN to predict target combinations and drug combinations for individual samples (DSCNi), DSCNi showed high correlation between target combinations predicted and real synergistic combinations (P = 1e-5) in pancreatic cell lines. In summary, DSCN is a highly effective computational method for the selection of target combinations.

Drug combinations have exhibited promising therapeutic effects in treating cancer patients with less toxicity and adverse side effects. However, it is infeasible to experimentally screen the enormous search space of all possible drug combinations. Therefore, developing computational models to efficiently and accurately identify potential anti-cancer synergistic drug combinations has attracted a lot of attention from the scientific community. Hypothesis-driven explicit mathematical methods or network pharmacology models have been popular in the last decade and have been comprehensively reviewed in previous surveys. With the surge of artificial intelligence and greater availability of large-scale datasets, machine learning especially deep learning methods are gaining popularity in the field of computational models for anti-cancer drug synergy prediction. Machine learning-based methods can be derived without strong assumptions about underlying mechanisms and have achieved state-of-the-art prediction performances, promoting much greater growth of the field. Here, we present a structured overview of available large-scale databases and machine learning especially deep learning methods in computational predictive models for anti-cancer drug synergy prediction. We provide a unified framework for machine learning models and detail existing model architectures as well as their contributions and limitations, shedding light into the future design of computational models. Besides, unbiased experiments are conducted to provide in-depth comparisons between reviewed papers in terms of their prediction performance.

Bayesian networks is a powerful method for identifying causal relationships among variables. However, as the network size increases, the time complexity of searching the optimal structure grows exponentially. We proposed a novel search algorithm - Fast and Furious Bayesian Network (FFBN). Compared to the existing greedy search algorithm, FFBN uses significantly fewer model configuration rules to determine the causal direction of edges when constructing the Bayesian network, which leads to greatly improved computational speed. We benchmarked the performance of FFBN by reconstructing gene regulatory networks (GRNs) from two DREAM5 challenge datasets: a synthetic dataset and a larger yeast transcriptome dataset. In both datasets, FFBN shows a much faster speed than the existing greedy search algorithm, while maintaining equally good or better performance in recall and precision. We then constructed three whole transcriptome GRNs for primary liver cancer (PL), primary colon cancer (PC) and colon to liver metastasis (CLM) expression data, which the existing greedy search algorithms failed. Three GRNs contain 12,099 common genes. Unprecedentedly, our newly developed FFBN algorithm is able to build up GRNs at a scale larger than 10,000 genes. Using FFBN, we discovered that CLM has its unique cancer molecular mechanisms and shares a certain degree of similarity with both PL and PC.

BACKGROUND: Cancer has been extensively characterized on the basis of genomics. The integration of genetic information about cancers with data on how the cancers respond to target based therapy to help to optimum cancer treatment. OBJECTIVE: The increasing usage of sequencing technology in cancer research and clinical practice has enormously advanced our understanding of cancer mechanisms. The cancer precision medicine is becoming a reality. Although off-label drug usage is a common practice in treating cancer, it suffers from the lack of knowledge base for proper cancer drug selections. This eminent need has become even more apparent considering the upcoming genomics data. METHODS: In this paper, a personalized medicine knowledge base is constructed by integrating various cancer drugs, drug-target database, and knowledge sources for the proper cancer drugs and their target selections. Based on the knowledge base, a bioinformatics approach for cancer drugs selection in precision medicine is developed. It integrates personal molecular profile data, including copy number variation, mutation, and gene expression. RESULTS: By analyzing the 85 triple negative breast cancer (TNBC) patient data in the Cancer Genome Altar, we have shown that 71.7% of the TNBC patients have FDA approved drug targets, and 51.7% of the patients have more than one drug target. Sixty-five drug targets are identified as TNBC treatment targets and 85 candidate drugs are recommended. Many existing TNBC candidate targets, such as Poly (ADP-Ribose) Polymerase 1 (PARP1), Cell division protein kinase 6 (CDK6), epidermal growth factor receptor, etc., were identified. On the other hand, we found some additional targets that are not yet fully investigated in the TNBC, such as Gamma-Glutamyl Hydrolase (GGH), Thymidylate Synthetase (TYMS), Protein Tyrosine Kinase 6 (PTK6), Topoisomerase (DNA) I, Mitochondrial (TOP1MT), Smoothened, Frizzled Class Receptor (SMO), etc. Our additional analysis of target and drug selection strategy is also fully supported by the drug screening data on TNBC cell lines in the Cancer Cell Line Encyclopedia. CONCLUSIONS: The proposed bioinformatics approach lays a foundation for cancer precision medicine. It supplies much needed knowledge base for the off-label cancer drug usage in clinics.

A new global nonlinear predictor with a particle swarm-optimized interval support vector regression (PSO-ISVR) is proposed to address three issues (viz., kernel selection, model optimization, kernel method speed) encountered when applying SVR in the presence of large data sets. The novel prediction model can reduce the SVR computing overhead by dividing input space and adaptively selecting the optimized kernel functions to obtain optimal SVR parameter by PSO. To quantify the quality of the predictor, its generalization performance and execution speed are investigated based on statistical learning theory. In addition, experiments using synthetic data as well as the stock volume weighted average price are reported to demonstrate the effectiveness of the developed models. The experimental results show that the proposed PSO-ISVR predictor can improve the computational efficiency and the overall prediction accuracy compared with the results produced by the SVR and other regression methods. The proposed PSO-ISVR provides an important tool for nonlinear regression analysis of big data.