Intelligent scoring system based on dynamic optical breast imaging for early detection of breast cancer.

Early detection of breast cancer can significantly improve patient outcomes and five-year survival in clinical screening. Dynamic optical breast imaging (DOBI) technology reflects the blood oxygen metabolism level of tumors based on the theory of tumor neovascularization, which offers a technical possibility for early detection of breast cancer. In this paper, we propose an intelligent scoring system integrating DOBI features assessment and a malignancy score grading reporting system for early detection of breast cancer. Specifically, we build six intelligent feature definition models to depict characteristics of regions of interest (ROIs) from location, space, time and context separately. Similar to the breast imaging-reporting and data system (BI-RADS), we conclude the malignancy score grading reporting system to score and evaluate ROIs as follows: Malignant (≥ 80 score), Likely Malignant (60-80 score), Intermediate (35-60 score), Likely Benign (10-35 score), and Benign (<10 score). This system eliminates the influence of subjective physician judgments on the assessment of the malignant probability of ROIs. Extensive experiments on 352 Chinese patients demonstrate the effectiveness of the proposed system compared to state-of-the-art methods.

LDSSNV: A Linkage Disequilibrium-Based Method for the Detection of Somatic Single-Nucleotide Variants.

Single nucleotide variants (SNVs) are very common in human genome and pose a significant effect on cellular proliferation and tumorigenesis in various cancers. Somatic variant and germline variant are the two forms of SNVs. They are the major drivers of inherited diseases and acquired tumors respectively. A reasonable analysis of the next generation sequencing data profiles from cancer genomes could provide crucial information for cancer diagnosis and treatment. Accurate detection of SNVs and distinguishing the two forms are still considered challenging tasks in cancer analysis. Herein, we propose a new approach, LDSSNV, to detect somatic SNVs without matched normal samples. LDSSNV predicts SNVs by training the XGboost classifier on a concise combination of features and distinguishes the two forms based on linkage disequilibrium which is a trait between germline mutations. LDSSNV provides two modes to distinguish the somatic variants from germline variants, the single-mode and multiple-mode by respectively using a single tumor sample and multiple tumor samples. The performance of the proposed method is assessed on both simulation data and real sequencing datasets. The analysis shows that the LDSSNV method outperforms competing methods and can become a robust and reliable tool for analyzing tumor genome variation.

svBreak: A New Approach for the Detection of Structural Variant Breakpoints Based on Convolutional Neural Network.

Structural variation (SV) is an important type of genome variation and confers susceptibility to human cancer diseases. Systematic analysis of SVs has become a crucial step for the exploration of mechanisms and precision diagnosis of cancers. The central point is how to accurately detect SV breakpoints by using next-generation sequencing (NGS) data. Due to the cooccurrence of multiple types of SVs in the human genome and the intrinsic complexity of SVs, the discrimination of SV breakpoint types is a challenging task. In this paper, we propose a convolutional neural network- (CNN-) based approach, called svBreak, for the detection and discrimination of common types of SV breakpoints. The principle of svBreak is that it extracts a set of SV-related features for each genome site from the sequencing reads aligned to the reference genome and establishes a data matrix where each row represents one site and each column represents one feature and then adopts a CNN model to analyze such data matrix for the prediction of SV breakpoints. The performance of the proposed approach is tested via simulation studies and application to a real sequencing sample. The experimental results demonstrate the merits of the proposed approach when compared with existing methods. Thus, svBreak can be expected to be a supplementary approach in the field of SV analysis in human tumor genomes.

A shortest path-based approach for copy number variation detection from next-generation sequencing data.

Copy number variation (CNV) is one of the main structural variations in the human genome and accounts for a considerable proportion of variations. As CNVs can directly or indirectly cause cancer, mental illness, and genetic disease in humans, their effective detection in humans is of great interest in the fields of oncogene discovery, clinical decision-making, bioinformatics, and drug discovery. The advent of next-generation sequencing data makes CNV detection possible, and a large number of CNV detection tools are based on next-generation sequencing data. Due to the complexity (e.g., bias, noise, alignment errors) of next-generation sequencing data and CNV structures, the accuracy of existing methods in detecting CNVs remains low. In this work, we design a new CNV detection approach, called shortest path-based Copy number variation (SPCNV), to improve the detection accuracy of CNVs. SPCNV calculates the k nearest neighbors of each read depth and defines the shortest path, shortest path relation, and shortest path cost sets based on which further calculates the mean shortest path cost of each read depth and its k nearest neighbors. We utilize the ratio between the mean shortest path cost for each read depth and the mean of the mean shortest path cost of its k nearest neighbors to construct a relative shortest path score formula that is able to determine a score for each read depth. Based on the score profile, a boxplot is then applied to predict CNVs. The performance of the proposed method is verified by simulation data experiments and compared against several popular methods of the same type. Experimental results show that the proposed method achieves the best balance between recall and precision in each set of simulated samples. To further verify the performance of the proposed method in real application scenarios, we then select real sample data from the 1,000 Genomes Project to conduct experiments. The proposed method achieves the best F1-scores in almost all samples. Therefore, the proposed method can be used as a more reliable tool for the routine detection of CNVs.

CIRCNV: Detection of CNVs Based on a Circular Profile of Read Depth from Sequencing Data.

Copy number variation (CNV) is a common type of structural variation in the human genome. Accurate detection of CNVs from tumor genomes can provide crucial information for the study of tumor genesis and cancer precision diagnosis. However, the contamination of normal genomes in tumor genomes and the crude profiles of the read depth make such a task difficult. In this paper, we propose an alternative approach, called CIRCNV, for the detection of CNVs from sequencing data. CIRCNV is an extension of our previously developed method CNV-LOF, which uses local outlier factors to predict CNVs. Comparatively, CIRCNV can be performed on individual tumor samples and has the following two new features: (1) it transfers the read depth profile from a line shape to a circular shape via a polar coordinate transformation, in order to improve the efficiency of the read depth (RD) profile for the detection of CNVs; and (2) it performs a second round of CNV declaration based on the truth circular RD profile, which is recovered by estimating tumor purity. We test and validate the performance of CIRCNV based on simulation and real sequencing data and perform comparisons with several peer methods. The results demonstrate that CIRCNV can obtain superior performance in terms of sensitivity and precision. We expect that our proposed method will be a supplement to existing methods and become a routine tool in the field of variation analysis of tumor genomes.

HBOS-CNV: A New Approach to Detect Copy Number Variations From Next-Generation Sequencing Data.

Copy number variation (CNV) is a genomic mutation that plays an important role in tumor evolution and tumor genesis. Accurate detection of CNVs from next-generation sequencing (NGS) data is still a challenging task due to artifacts such as uneven mapped reads and unbalanced amplitudes of gains and losses. This study proposes a new approach called HBOS-CNV to detect CNVs from NGS data. The central point of HBOS-CNV is that it uses a new statistic, the histogram-based outlier score (HBOS), to evaluate the fluctuation of genome bins to determine those of changed copy numbers. In comparison with existing statistics in the evaluation of CNVs, HBOS is a non-linearly transformed value from the observed read depth (RD) value of each genome bin, having the potential ability to relieve the effects resulted from the above artifacts. In the calculation of HBOS values, a dynamic width histogram is utilized to depict the density of bins on the genome being analyzed, which can reduce the effects of noises partially contributed by mapping and sequencing errors. The evaluation of genome bins using such a new statistic can lead to less extremely significant CNVs having a high probability of detection. We evaluated this method using a large number of simulation datasets and compared it with four existing methods (CNVnator, CNV-IFTV, CNV-LOF, and iCopyDav). The results demonstrated that our proposed method outperforms the others in terms of sensitivity, precision, and F1-measure. Furthermore, we applied the proposed method to a set of real sequencing samples from the 1000 Genomes Project and determined a number of CNVs with biological meanings. Thus, the proposed method can be regarded as a routine approach in the field of genome mutation analysis for cancer samples.

A novel machine learning approach (svmSomatic) to distinguish somatic and germline mutations using next-generation sequencing data.

Somatic mutations are a large category of genetic variations, which play an essential role in tumorigenesis. Detection of somatic single nucleotide variants (SNVs) could facilitate downstream analysis of tumorigenesis. Many computational methods have been developed to detect SNVs, but most require normal matched samples to differentiate somatic SNVs from the normal state, which can be difficult to obtain. Therefore, developing new approaches for detecting somatic SNVs without matched samples are crucial. In this work, we detected somatic mutations from individual tumor samples based on a novel machine learning approach, svmSomatic, using next-generation sequencing (NGS) data. In addition, as somatic SNV detection can be impacted by multiple mutations, with germline mutations and co-occurrence of copy number variations (CNVs) common in organisms, we used the novel approach to distinguish somatic and germline mutations based on the NGS data from individual tumor samples. In summary, svmSomatic: (1) considers the influence of CNV co-occurrence in detecting somatic mutations; and (2) trains a support vector machine algorithm to distinguish between somatic and germline mutations, without requiring normal matched samples. We further tested and compared svmSomatic with other common methods. Results showed that svmSomatic performance, as measured by F1-score, was significantly better than that of others using both simulation and real NGS data.

KNNCNV: A K-Nearest Neighbor Based Method for Detection of Copy Number Variations Using NGS Data.

Copy number variation (CNV) is a well-known type of genomic mutation that is associated with the development of human cancer diseases. Detection of CNVs from the human genome is a crucial step for the pipeline of starting from mutation analysis to cancer disease diagnosis and treatment. Next-generation sequencing (NGS) data provides an unprecedented opportunity for CNVs detection at the base-level resolution, and currently, many methods have been developed for CNVs detection using NGS data. However, due to the intrinsic complexity of CNVs structures and NGS data itself, accurate detection of CNVs still faces many challenges. In this paper, we present an alternative method, called KNNCNV (K-Nearest Neighbor based CNV detection), for the detection of CNVs using NGS data. Compared to current methods, KNNCNV has several distinctive features: 1) it assigns an outlier score to each genome segment based solely on its first k nearest-neighbor distances, which is not only easy to extend to other data types but also improves the power of discovering CNVs, especially the local CNVs that are likely to be masked by their surrounding regions; 2) it employs the variational Bayesian Gaussian mixture model (VBGMM) to transform these scores into a series of binary labels without a user-defined threshold. To evaluate the performance of KNNCNV, we conduct both simulation and real sequencing data experiments and make comparisons with peer methods. The experimental results show that KNNCNV could derive better performance than others in terms of F1-score.

STIC: Predicting Single Nucleotide Variants and Tumor Purity in Cancer Genome.

Single nucleotide variant (SNV) plays an important role in cellular proliferation and tumorigenesis in various types of human cancer. Next-generation sequencing (NGS) has provided high-throughput data at an unprecedented resolution to predict SNVs. Currently, there exist many computational methods for either germline or somatic SNV discovery from NGS data, but very few of them are versatile enough to adapt to any situations. In the absence of matched normal samples, the prediction of somatic SNVs from single-tumor samples becomes considerably challenging, especially when the tumor purity is unknown. Here, we propose a new approach, STIC, to predict somatic SNVs and estimate tumor purity from NGS data without matched normal samples. The main features of STIC include: (1) extracting a set of SNV-relevant features on each site and training the BP neural network algorithm on the features to predict SNVs; (2) creating an iterative process to distinguish somatic SNVs from germline ones by disturbing allele frequency; and (3) establishing a reasonable relationship between tumor purity and allele frequencies of somatic SNVs to accurately estimate the purity. We quantitatively evaluate the performance of STIC on both simulation and real sequencing datasets, the results of which indicate that STIC outperforms competing methods.

CNV_IFTV: An Isolation Forest and Total Variation-Based Detection of CNVs from Short-Read Sequencing Data.

Accurate detection of copy number variations (CNVs) from short-read sequencing data is challenging due to the uneven distribution of reads and the unbalanced amplitudes of gains and losses. The direct use of read depths to measure CNVs tends to limit performance. Thus, robust computational approaches equipped with appropriate statistics are required to detect CNV regions and boundaries. This study proposes a new method called CNV_IFTV to address this need. CNV_IFTV assigns an anomaly score to each genome bin through a collection of isolation trees. The trees are trained based on isolation forest algorithm through conducting subsampling from measured read depths. With the anomaly scores, CNV_IFTV uses a total variation model to smooth adjacent bins, leading to a denoised score profile. Finally, a statistical model is established to test the denoised scores for calling CNVs. CNV_IFTV is tested on both simulated and real data in comparison to several peer methods. The results indicate that the proposed method outperforms the peer methods. CNV_IFTV is a reliable tool for detecting CNVs from short-read sequencing data even for low-level coverage and tumor purity. The detection results on tumor samples can aid to evaluate known cancer genes and to predict target drugs for disease diagnosis.

RKDOSCNV: A Local Kernel Density-Based Approach to the Detection of Copy Number Variations by Using Next-Generation Sequencing Data.

Copy number variations (CNVs) are significant causes of many human cancers and genetic diseases. The detection of CNVs has become a common method by which to analyze human diseases using next-generation sequencing (NGS) data. However, effective detection of insignificant CNVs is still a challenging task. In this study, we propose a new detection method, RKDOSCNV, to meet the need. RKDOSCNV uses kernel density estimation method to evaluate the local kernel density distribution of each read depth segment (RDS) based on an expanded nearest neighbor (k-nearest neighbors, reverse nearest neighbors, and shared nearest neighbors of each RDS) data set, and assigns a relative kernel density outlier score (RKDOS) for each RDS. According to the RKDOS profile, RKDOSCNV predicts the candidate CNVs by choosing a reasonable threshold, which it uses split read approach to correct the boundaries of candidate CNVs. The performance of RKDOSCNV is assessed by comparing it with several current popular methods via experiments with simulated and real data at different tumor purity levels. The experimental results verify that the performance of RKDOSCNV is superior to that of several other methods. In summary, RKDOSCNV is a simple and effective method for the detection of CNVs from whole genome sequencing (WGS) data, especially for samples with low tumor purity.

DINTD: Detection and Inference of Tandem Duplications From Short Sequencing Reads.

Tandem duplication (TD) is an important type of structural variation (SV) in the human genome and has biological significance for human cancer evolution and tumor genesis. Accurate and reliable detection of TDs plays an important role in advancing early detection, diagnosis, and treatment of disease. The advent of next-generation sequencing technologies has made it possible for the study of TDs. However, detection is still challenging due to the uneven distribution of reads and the uncertain amplitude of TD regions. In this paper, we present a new method, DINTD (Detection and INference of Tandem Duplications), to detect and infer TDs using short sequencing reads. The major principle of the proposed method is that it first extracts read depth and mapping quality signals, then uses the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm to find the possible TD regions. The total variation penalized least squares model is fitted with read depth and mapping quality signals to denoise signals. A 2D binary search tree is used to search the neighbor points effectively. To further identify the exact breakpoints of the TD regions, split-read signals are integrated into DINTD. The experimental results of DINTD on simulated data sets showed that DINTD can outperform other methods for sensitivity, precision, F1-score, and boundary bias. DINTD is further validated on real samples, and the experiment results indicate that it is consistent with other methods. This study indicates that DINTD can be used as an effective tool for detecting TDs.

MFCNV: A New Method to Detect Copy Number Variations From Next-Generation Sequencing Data.

Copy number variation (CNV) is a very important phenomenon in tumor genomes and plays a significant role in tumor genesis. Accurate detection of CNVs has become a routine and necessary procedure for a deep investigation of tumor cells and diagnosis of tumor patients. Next-generation sequencing (NGS) technique has provided a wealth of data for the detection of CNVs at base-pair resolution. However, such task is usually influenced by a number of factors, including GC-content bias, sequencing errors, and correlations among adjacent positions within CNVs. Although many existing methods have dealt with some of these artifacts by designing their own strategies, there is still a lack of comprehensive consideration of all the factors. In this paper, we propose a new method, MFCNV, for an accurate detection of CNVs from NGS data. Compared with existing methods, the characteristics of the proposed method include the following: (1) it makes a full consideration of the intrinsic correlations among adjacent positions in the genome to be analyzed, (2) it calculates read depth, GC-content bias, base quality, and correlation value for each genome bin and combines them as multiple features for the evaluation of genome bins, and (3) it addresses the joint effect among the factors via training a neural network algorithm for the prediction of CNVs. We test the performance of the MFCNV method by using simulation and real sequencing data and make comparisons with several peer methods. The results demonstrate that our method is superior to other methods in terms of sensitivity, precision, and F1-score and can detect many CNVs that other methods have not discovered. MFCNV is expected to be a complementary tool in the analysis of mutations in tumor genomes and can be extended to be applied to the analysis of single-cell sequencing data.

Accurate Inference of Tumor Purity and Absolute Copy Numbers From High-Throughput Sequencing Data.

Inference of absolute copy numbers in tumor genomes is one of the key points in the study of tumor genesis. However, the mixture of tumor and normal cells poses a big challenge to this task. Accurate estimation of tumor purity (i.e., the fraction of tumor cells) is a necessary step to solve this problem. In this paper, we propose a new approach, AITAC, to accurately infer tumor purity and absolute copy numbers in a tumor sample by using high-throughput sequencing (HTS) data. In contrast to many existing algorithms for estimating tumor purity, which usually rely on pre-detected mutation genotypes (heterogeneity and homogeneity), AITAC just requires read depths (RDs) observed at the regions with copy number losses. AITAC creates a non-linear model to correlate tumor purity, observed and expected RDs. It adopts an exhaustive search strategy to scan tumor purity in a wide range, and chooses the tumor purity that minimizes the deviation between observed RDs and expected ones as the optimal solution. We apply the proposed approach to both simulation and real sequencing data sets and demonstrate its performance by comparing with two classical approaches. AITAC is freely available at https://github.com/BDanalysis/aitac and can be expected to become a useful approach for researchers to analyze copy numbers in cancer genome.

CONDEL: Detecting Copy Number Variation and Genotyping Deletion Zygosity from Single Tumor Samples Using Sequence Data.

Characterizing copy number variations (CNVs) from sequenced genomes is a both feasible and cost-effective way to search for driver genes in cancer diagnosis. A number of existing algorithms for CNV detection only explored part of the features underlying sequence data and copy number structures, resulting in limited performance. Here, we describe CONDEL, a method for detecting CNVs from single tumor samples using high-throughput sequence data. CONDEL utilizes a novel statistic in combination with a peel-off scheme to assess the statistical significance of genome bins, and adopts a Bayesian approach to infer copy number gains, losses, and deletion zygosity based on statistical mixture models. We compare CONDEL to six peer methods on a large number of simulation datasets, showing improved performance in terms of true positive and false positive rates, and further validate CONDEL on three real datasets derived from the 1000 Genomes Project and the EGA archive. CONDEL obtained higher consistent results in comparison with other three single sample-based methods, and exclusively identified a number of CNVs that were previously associated with cancers. We conclude that CONDEL is a powerful tool for detecting copy number variations on single tumor samples even if these are sequenced at low-coverage.

SVSR: A Program to Simulate Structural Variations and Generate Sequencing Reads for Multiple Platforms.

Structural variation accounts for a major fraction of mutations in the human genome and confers susceptibility to complex diseases. Next generation sequencing along with the rapid development of computational methods provides a cost-effective procedure to detect such variations. Simulation of structural variations and sequencing reads with real characteristics is essential for benchmarking the computational methods. Here, we develop a new program, SVSR, to simulate five types of structural variations (indels, tandem duplication, CNVs, inversions, and translocations) and SNPs for the human genome and to generate sequencing reads with features from popular platforms (Illumina, SOLiD, 454, and Ion Torrent). We adopt a selection model trained from real data to predict copy number states, starting from the first site of a particular genome to the end. Furthermore, we utilize references of microbial genomes to produce insertion fragments and design probabilistic models to imitate inversions and translocations. Moreover, we create platform-specific errors and base quality profiles to generate normal, tumor, or normal-tumor mixture reads. Experimental results show that SVSR could capture more features that are realistic and generate datasets with satisfactory quality scores. SVSR is able to evaluate the performance of structural variation detection methods and guide the development of new computational methods.

Inferring subgroup-specific driver genes from heterogeneous cancer samples via subspace learning with subgroup indication.

MOTIVATION: Detecting driver genes from gene mutation data is a fundamental task for tumorigenesis research. Due to the fact that cancer is a heterogeneous disease with various subgroups, subgroup-specific driver genes are the key factors in the development of precision medicine for heterogeneous cancer. However, the existing driver gene detection methods are not designed to identify subgroup specificities of their detected driver genes, and therefore cannot indicate which group of patients is associated with the detected driver genes, which is difficult to provide specifically clinical guidance for individual patients. RESULTS: By incorporating the subspace learning framework, we propose a novel bioinformatics method called DriverSub, which can efficiently predict subgroup-specific driver genes in the situation where the subgroup annotations are not available. When evaluated by simulation datasets with known ground truth and compared with existing methods, DriverSub yields the best prediction of driver genes and the inference of their related subgroups. When we apply DriverSub on the mutation data of real heterogeneous cancers, we can observe that the predicted results of DriverSub are highly enriched for experimentally validated known driver genes. Moreover, the subgroups inferred by DriverSub are significantly associated with the annotated molecular subgroups, indicating its capability of predicting subgroup-specific driver genes. AVAILABILITY AND IMPLEMENTATION: The source code is publicly available at https://github.com/JianingXi/DriverSub. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Integrating Somatic Mutations for Breast Cancer Survival Prediction Using Machine Learning Methods.

Breast cancer is the most common malignancy in women, and because it has a high mortality rate, it is urgent to develop computational methods to increase the accuracy of breast cancer survival predictive models. Although multi-omics data such as gene expression have been extensively used in recent studies, the accurate prognosis of breast cancer remains a challenge. Somatic mutations are another important and promising data source for studying cancer development, and its effect on the prognosis of breast cancer remains to be further explored. Meanwhile, these omics datasets are high-dimensional and redundant. Therefore, we adopted multiple kernel learning (MKL) to efficiently integrate somatic mutation to currently molecular data including gene expression, copy number variation (CNV), methylation, and protein expression data for the prediction of breast cancer survival. Before integration, the maximum relevance minimum redundancy (mRMR) feature selection method was utilized to select features that present high relevance to survival and low redundancy among themselves for each type of data. The experimental results demonstrated that the proposed method achieved the most optimal performance and there was a remarkable improvement in the prediction performance when somatic mutations were included, indicating that somatic mutations are critical for improving breast cancer survival predictions. Moreover, mRMR was superior to other feature selection methods used in previous studies. Furthermore, MKL outperformed the other traditional classifiers in multi-omics data integration. Our analysis indicated that through employing promising omics data such as somatic mutations and harnessing the power of proper feature selection methods and effective integration frameworks, the breast cancer survival predictive accuracy can be further increased, thereby providing a more optimal clinical diagnosis and more effective treatment for breast cancer patients.

Stratification of Breast Cancer by Integrating Gene Expression Data and Clinical Variables.

Breast cancer is a heterogeneous disease. Although gene expression profiling has led to the definition of several subtypes of breast cancer, the precise discovery of the subtypes remains a challenge. Clinical data is another promising source. In this study, clinical variables are utilized and integrated to gene expressions for the stratification of breast cancer. We adopt two phases: gene selection and clustering, where the integration is in the gene selection phase; only genes whose expressions are most relevant to each clinical variable and least redundant among themselves are selected for further clustering. In practice, we simply utilize maximum relevance minimum redundancy (mRMR) for gene selection and k-means for clustering. We compare the results of our method with those of two commonly used only expression-based breast cancer stratification methods: prediction analysis of microarray 50 (PAM50) and highest variability (HV). The result is that our method outperforms them in identifying subtypes significantly associated with five-year survival and recurrence time. Specifically, our method identified recurrence-associated breast cancer subtypes that were not identified by PAM50 and HV. Additionally, our analysis discovered three survival-associated luminal-A subgroups and two survival-associated luminal-B subgroups. The study indicates that screening clinically relevant gene expressions yields improved breast cancer stratification.