Traffic classification is crucial for network management and security. Recently, deep learning-based methods have demonstrated good performance in traffic classification. However, they primarily capture features from raw packet bytes, overlooking the significance of inter-packet correlations within flows from a global perspective. Additionally, effectively handling both packet-length and temporal information, while extracting the structural relationships from a graph into the model, remains a challenge for enhancing the performance of traffic prediction. In this paper, we propose DigTraffic, a novel dual-channel interactive graph transformer to address these limitations. DigTraffic employs a message-level graph-structured flow representation combined with message-aware structural aggregation. To learn intrinsic flow representations, DigTraffic constructs traffic interaction graphs, by incorporating three well-designed heterogeneous types of edges to capture client-server interactions. After that, we separately encode lengthy and temporal flow sequences using a dual-channel network and fuse these modalities within a Transformer architecture. Furthermore, DigTraffic introduces a message-aware Graph Transformer that leverages both node embeddings and edge spatial relations to capture complex graph structures and rich structural information. Experimental results demonstrate that our method significantly outperforms the state-of-the-art methods on four real-world traffic datasets.

Although traditional intrusion detection systems can identify unknown attacks based on established patterns, they are incapable of the fine-grained classification needed to distinguish one from another. Moreover, the prohibitive cost of adapting a detection model to new threats without degrading its existing capabilities severely impedes its effectiveness against emerging attacks. To address these challenges, we propose a Multi-View Learning Framework for Class-Incremental Unknown Attack Detection (MVL-AD). This framework designs a contrastive error autoencoder (CE-AE) for each class to build multiple differentiated learners. These learners calculate anomaly scores from diverse perspectives to comprehensively characterize the attack profile. We apply anomaly detection algorithms to the attack profiles to identify unknown attacks. These identified unknown attacks are classified into distinct categories using Mini-Batch K-Means, with the optimal number of clusters determined by the Calinski-Harabasz (CH) index. Furthermore, MVL-AD incorporates an adaptive updating mechanism to improve its detection performance on both known and unknown attacks. Extensive experiments demonstrate that MVL-AD outperforms state-of-the-art methods in stability, scalability, and practicability.

MOTIVATION: Flow cytometry (FCM) and transcription profiling are the two widely used assays in translational immunology research. However, there is no data integration pipeline for analyzing these two types of assays together with experiment variables for biomarker inference. Current FCM data analysis mainly relies on subjective manual gating analysis, which is difficult to be directly integrated with other automated computational methods. Existing deconvolutional analysis of bulk transcriptomics relies on predefined marker genes in the transcriptomics data, which are unavailable for novel cell types and does not utilize the FCM data that provide canonical phenotypic definitions of the cell types. RESULTS: We developed a novel analytics pipeline-FastMix-for computational immunology, which integrates flow cytometry, bulk transcriptomics and clinical covariates for identifying cell type-specific gene expression signatures and biomarker genes. FastMix addresses the 'large p, small n' problem in the gene expression and flow cytometry integration analysis via a linear mixed effects model (LMER) for both cross-sectional and longitudinal studies. Its novel moment-based estimator not only reduces bias in parameter estimation but also is more efficient than iterative optimization. The FastMix pipeline also includes a cutting-edge flow cytometry data analysis method-DAFi-for identifying cell populations of interest and their characteristics. Simulation studies showed that FastMix produced smaller type I/II errors than competing methods. Validation using real data of two vaccine studies showed that FastMix identified a consistent set of signature genes as in independent single-cell RNA-seq analysis, producing additional interesting findings. AVAILABILITY AND IMPLEMENTATION: Source code of FastMix is publicly available at https://github.com/terrysun0302/FastMix. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: We developed super-delta2, a differential gene expression analysis pipeline designed for multi-group comparisons for RNA-seq data. It includes a customized one-way ANOVA F-test and a post-hoc test for pairwise group comparisons; both are designed to work with a multivariate normalization procedure to reduce technical noise. It also includes a trimming procedure with bias-correction to obtain robust and approximately unbiased summary statistics used in these tests. We demonstrated the asymptotic applicability of super-delta2 to log-transformed read counts in RNA-seq data by large sample theory based on Negative Binomial Poisson (NBP) distribution. RESULTS: We compared super-delta2 with three commonly used RNA-seq data analysis methods: limma/voom, edgeR and DESeq2 using both simulated and real datasets. In all three simulation settings, super-delta2 not only achieved the best overall statistical power, but also was the only method that controlled type I error at the nominal level. When applied to a breast cancer dataset to identify differential expression pattern associated with multiple pathologic stages, super-delta2 selected more enriched pathways than other methods, which are directly linked to the underlying biological condition (breast cancer). CONCLUSIONS: In conclusion, by incorporating trimming and bias-correction in the normalization step, super-delta2 was able to achieve tight control of type I error. Because the hypothesis tests are based on asymptotic normal approximation of the NBP distribution, super-delta2 does not require computationally expensive iterative optimization procedures used by methods such as edgeR and DESeq2, which occasionally have convergence issues. AVAILABILITY AND IMPLEMENTATION: Our method is implemented in a R-package, 'superdelta2', freely available at: https://github.com/fhlsjs/superdelta2. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Respiratory syncytial virus (RSV) infection results in millions of hospitalizations and thousands of deaths each year. Variations in the adaptive and innate immune response appear to be associated with RSV severity. To investigate the host response to RSV infection in infants, we performed a systems-level study of RSV pathophysiology, incorporating high-throughput measurements of the peripheral innate and adaptive immune systems and the airway epithelium and microbiota. We implemented a novel multi-omic data integration method based on multilayered principal component analysis, penalized regression, and feature weight back-propagation, which enabled us to identify cellular pathways associated with RSV severity. In both airway and immune cells, we found an association between RSV severity and activation of pathways controlling Th17 and acute phase response signaling, as well as inhibition of B cell receptor signaling. Dysregulation of both the humoral and mucosal response to RSV may play a critical role in determining illness severity.

MOTIVATION: Combining gene expression (GE) profiles generated from different platforms enables previously infeasible studies due to sample size limitations. Several cross-platform normalization methods have been developed to remove the systematic differences between platforms, but they may also remove meaningful biological differences among datasets. In this work, we propose a novel approach that removes the platform, not the biological differences. Dubbed as 'MatchMixeR', we model platform differences by a linear mixed effects regression (LMER) model, and estimate them from matched GE profiles of the same cell line or tissue measured on different platforms. The resulting model can then be used to remove platform differences in other datasets. By using LMER, we achieve better bias-variance trade-off in parameter estimation. We also design a computationally efficient algorithm based on the moment method, which is ideal for ultra-high-dimensional LMER analysis. RESULTS: Compared with several prominent competing methods, MatchMixeR achieved the highest after-normalization concordance. Subsequent differential expression analyses based on datasets integrated from different platforms showed that using MatchMixeR achieved the best trade-off between true and false discoveries, and this advantage is more apparent in datasets with limited samples or unbalanced group proportions. AVAILABILITY AND IMPLEMENTATION: Our method is implemented in a R-package, 'MatchMixeR', freely available at: https://github.com/dy16b/Cross-Platform-Normalization. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: For many practical hypothesis testing (H-T) applications, the data are correlated and/or with heterogeneous variance structure. The regression t-test for weighted linear mixed-effects regression (LMER) is a legitimate choice because it accounts for complex covariance structure; however, high computational costs and occasional convergence issues make it impractical for analyzing high-throughput data. In this paper, we propose computationally efficient parametric and semiparametric tests based on a set of specialized matrix techniques dubbed as the PB-transformation. The PB-transformation has two advantages: 1. The PB-transformed data will have a scalar variance-covariance matrix. 2. The original H-T problem will be reduced to an equivalent one-sample H-T problem. The transformed problem can then be approached by either the one-sample Student's t-test or Wilcoxon signed rank test. RESULTS: In simulation studies, the proposed methods outperform commonly used alternative methods under both normal and double exponential distributions. In particular, the PB-transformed t-test produces notably better results than the weighted LMER test, especially in the high correlation case, using only a small fraction of computational cost (3 versus 933 s). We apply these two methods to a set of RNA-seq gene expression data collected in a breast cancer study. Pathway analyses show that the PB-transformed t-test reveals more biologically relevant findings in relation to breast cancer than the weighted LMER test. CONCLUSIONS: As fast and numerically stable replacements for the weighted LMER test, the PB-transformed tests are especially suitable for "messy" high-throughput data that include both independent and matched/repeated samples. By using our method, the practitioners no longer have to choose between using partial data (applying paired tests to only the matched samples) or ignoring the correlation in the data (applying two sample tests to data with some correlated samples). Our method is implemented as an R package 'PBtest' and is available at https://github.com/yunzhang813/PBtest-R-Package .

MOTIVATION: Gene set enrichment analyses (GSEAs) are widely used in genomic research to identify underlying biological mechanisms (defined by the gene sets), such as Gene Ontology terms and molecular pathways. There are two caveats in the currently available methods: (i) they are typically designed for group comparisons or regression analyses, which do not utilize temporal information efficiently in time-series of transcriptomics measurements; and (ii) genes overlapping in multiple molecular pathways are considered multiple times in hypothesis testing. RESULTS: We propose an inferential framework for GSEA based on functional data analysis, which utilizes the temporal information based on functional principal component analysis, and disentangles the effects of overlapping genes by a functional extension of the elastic-net regression. Furthermore, the hypothesis testing for the gene sets is performed by an extension of Mann-Whitney U test which is based on weighted rank sums computed from correlated observations. By using both simulated datasets and a large-scale time-course gene expression data on human influenza infection, we demonstrate that our method has uniformly better receiver operating characteristic curves, and identifies more pathways relevant to immune-response to human influenza infection than the competing approaches. AVAILABILITY AND IMPLEMENTATION: The methods are implemented in R package FUNNEL, freely and publicly available at: https://github.com/yunzhang813/FUNNEL-GSEA-R-Package . CONTACT: [email protected] or [email protected]. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: Normalization is an important data preparation step in gene expression analyses, designed to remove various systematic noise. Sample variance is greatly reduced after normalization, hence the power of subsequent statistical analyses is likely to increase. On the other hand, variance reduction is made possible by borrowing information across all genes, including differentially expressed genes (DEGs) and outliers, which will inevitably introduce some bias. This bias typically inflates type I error; and can reduce statistical power in certain situations. In this study we propose a new differential expression analysis pipeline, dubbed as super-delta, that consists of a multivariate extension of the global normalization and a modified t-test. A robust procedure is designed to minimize the bias introduced by DEGs in the normalization step. The modified t-test is derived based on asymptotic theory for hypothesis testing that suitably pairs with the proposed robust normalization. RESULTS: We first compared super-delta with four commonly used normalization methods: global, median-IQR, quantile, and cyclic loess normalization in simulation studies. Super-delta was shown to have better statistical power with tighter control of type I error rate than its competitors. In many cases, the performance of super-delta is close to that of an oracle test in which datasets without technical noise were used. We then applied all methods to a collection of gene expression datasets on breast cancer patients who received neoadjuvant chemotherapy. While there is a substantial overlap of the DEGs identified by all of them, super-delta were able to identify comparatively more DEGs than its competitors. Downstream gene set enrichment analysis confirmed that all these methods selected largely consistent pathways. Detailed investigations on the relatively small differences showed that pathways identified by super-delta have better connections to breast cancer than other methods. CONCLUSIONS: As a new pipeline, super-delta provides new insights to the area of differential gene expression analysis. Solid theoretical foundation supports its asymptotic unbiasedness and technical noise-free properties. Implementation on real and simulated datasets demonstrates its decent performance compared with state-of-art procedures. It also has the potential of expansion to be incorporated with other data type and/or more general between-group comparison problems.

BACKGROUND: Quantile and rank normalizations are two widely used pre-processing techniques designed to remove technological noise presented in genomic data. Subsequent statistical analysis such as gene differential expression analysis is usually based on normalized expressions. In this study, we find that these normalization procedures can have a profound impact on differential expression analysis, especially in terms of testing power. RESULTS: We conduct theoretical derivations to show that the testing power of differential expression analysis based on quantile or rank normalized gene expressions can never reach 100% with fixed sample size no matter how strong the gene differentiation effects are. We perform extensive simulation analyses and find the results corroborate theoretical predictions. CONCLUSIONS: Our finding may explain why genes with well documented strong differentiation are not always detected in microarray analysis. It provides new insights in microarray experimental design and will help practitioners in selecting proper normalization procedures.

BACKGROUND: Microarray gene differential expression analysis is a widely used technique that deals with high dimensional data and is computationally intensive for permutation-based procedures. Microarray gene differential association analysis is even more computationally demanding and must take advantage of multicore computing technology, which is the driving force behind increasing compute power in recent years. In this paper, we present a two-layer hierarchical parallel implementation of gene differential association analysis. It takes advantage of both fine- and coarse-grain (with granularity defined by the frequency of communication) parallelism in order to effectively leverage the non-uniform nature of parallel processing available in the cutting-edge systems of today. RESULTS: Our results show that this hierarchical strategy matches data sharing behavior to the properties of the underlying hardware, thereby reducing the memory and bandwidth needs of the application. The resulting improved efficiency reduces computation time and allows the gene differential association analysis code to scale its execution with the number of processors. The code and biological data used in this study are downloadable from http://www.urmc.rochester.edu/biostat/people/faculty/hu.cfm. CONCLUSIONS: The performance sweet spot occurs when using a number of threads per MPI process that allows the working sets of the corresponding MPI processes running on the multicore to fit within the machine cache. Hence, we suggest that practitioners follow this principle in selecting the appropriate number of MPI processes and threads within each MPI process for their cluster configurations. We believe that the principles of this hierarchical approach to parallelization can be utilized in the parallelization of other computationally demanding kernels.

MOTIVATION: Very little attention has been given to gene selection procedures based on intergene correlation structure, which is often neglected in the context of differential gene expression analysis. We propose a statistical procedure to select genes that have different associations with others across different phenotypes. This procedure is based on a new gene association score, called the covariance distance. RESULTS: We apply the proposed method, along with two alternative methods, to several simulated datasets and find out that our method is much more powerful than the other two. For biological data, we demonstrate that the analysis of differentially associated genes complements the analysis of differentially expressed genes. Combining both procedures provides a more comprehensive functional interpretation of the experimental results. AVAILABILITY: The code is downloadable from http://www.urmc.rochester.edu/biostat/people/faculty/hu.cfm. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: Microarray technology is commonly used as a simple screening tool with a focus on selecting genes that exhibit extremely large differential expressions between different phenotypes. It lacks the ability to select genes that change their relationships with other genes in different biological conditions (differentially correlated genes). We intend to enrich the above procedure by proposing a nonparametric selection procedure that selects differentially correlated genes. RESULTS: Using both simulations and resampling techniques, we found that our procedure correctly detected genes that were not differentially expressed but differentially correlated. We also applied our procedure to a set of biological data and found some potentially important genes that were not selected by the traditional method. DISCUSSION AND CONCLUSION: Microarray technology yields multidimensional information on the function of the whole genome. Rather than treating intergene correlation as a nuisance to the traditional gene selection procedures which are essentially univariate, our method utilizes the rich information contained in the correlation as a new selection criterion. It can provide additional useful candidate genes for the biologists.

BACKGROUND: The number of genes declared differentially expressed is a random variable and its variability can be assessed by resampling techniques. Another important stability indicator is the frequency with which a given gene is selected across subsamples. We have conducted studies to assess stability and some other properties of several gene selection procedures with biological and simulated data. RESULTS: Using resampling techniques we have found that some genes are selected much less frequently (across sub-samples) than other genes with the same adjusted p-values. The extent to which this type of instability manifests itself can be assessed by a method introduced in this paper. The effect of correlation between gene expression levels on the performance of multiple testing procedures is studied by computer simulations. CONCLUSION: Resampling represents a tool for reducing the set of initially selected genes to those with a sufficiently high selection frequency. Using resampling techniques it is also possible to assess variability of different performance indicators. Stability properties of several multiple testing procedures are described at length in the present paper.

BACKGROUND: Stochastic dependence between gene expression levels in microarray data is of critical importance for the methods of statistical inference that resort to pooling test-statistics across genes. It is frequently assumed that dependence between genes (or tests) is sufficiently weak to justify the proposed methods of testing for differentially expressed genes. A potential impact of between-gene correlations on the performance of such methods has yet to be explored. RESULTS: The paper presents a systematic study of correlation between the t-statistics associated with different genes. We report the effects of four different normalization methods using a large set of microarray data on childhood leukemia in addition to several sets of simulated data. Our findings help decipher the correlation structure of microarray data before and after the application of normalization procedures. CONCLUSION: A long-range correlation in microarray data manifests itself in thousands of genes that are heavily correlated with a given gene in terms of the associated t-statistics. By using normalization methods it is possible to significantly reduce correlation between the t-statistics computed for different genes. Normalization procedures affect both the true correlation, stemming from gene interactions, and the spurious correlation induced by random noise. When analyzing real world biological data sets, normalization procedures are unable to completely remove correlation between the test statistics. The long-range correlation structure also persists in normalized data.