PenDA, a rank-based method for personalized differential analysis: Application to lung cancer

Personalized analysis of human cancer multi-omics for precision oncology

Multi-omics technologies, encompassing genomics, proteomics, and transcriptomics, provide profound insights into cancer biology. A fundamental computational approach for analyzing multi-omics data is differential analysis, which identifies molecular distinctions between cancerous and normal tissues. Traditional methods, however, often fail to address the distinct heterogeneity of individual tumors, thereby neglecting crucial patient-specific molecular traits. This shortcoming underscores the necessity for tailored differential analysis algorithms, which focus on particular patient variations. Such approaches offer a more nuanced understanding of cancer biology and are instrumental in pinpointing personalized therapeutic strategies. In this review, we summarize the principles of current individualized techniques. We also review their efficacy in analyzing cancer multi-omics data and discuss their potential applications in clinical practice.

Personalized differential expression analysis in triple-negative breast cancer

Identification of individual-level differentially expressed genes (DEGs) is a pre-step for the analysis of disease-specific biological mechanisms and precision medicine. Previous algorithms cannot balance accuracy and sufficient statistical power. Herein, RankCompV2, designed for identifying population-level DEGs based on relative expression orderings, was adjusted to identify individual-level DEGs. Furthermore, an optimized version of individual-level RankCompV2, named as RankCompV2.1, was designed based on the assumption that the rank positions of genes and relative rank differences of gene pairs would influence the identification of individual-level DEGs. In comparison to other individualized analysis algorithms, RankCompV2.1 performed better on statistical power, computational efficiency, and acquired coequal accuracy in both simulation and real paired cancer-normal data from ten cancer types. Besides, single sample GSEA and Gene Set Variation Analysis analysis showed that pathways enriched with up-regulated and down-regulated genes presented higher and lower enrichment scores, respectively. Furthermore, we identified 16 genes that were universally deregulated in 966 triple-negative breast cancer (TNBC) samples and interacted with Food and Drug Administration (FDA)-approved drugs or antineoplastic agents, indicating notable therapeutic targets for TNBC. In addition, we also identified genes with highly variable deregulation status and used these genes to cluster TNBC samples into three subgroups with different prognoses. The subgroup with the poorest outcome was characterized by down-regulated immune-regulated pathways, signal transduction pathways, and apoptosis-related pathways. Protein-protein interaction network analysis revealed that OAS family genes may be promising drug targets to activate tumor immunity in this subgroup. In conclusion, RankCompV2.1 is capable of identifying individual-level DEGs with high accuracy and statistical power, analyzing mechanisms of carcinogenesis and exploring therapeutic strategy.

Identifying cancer driver genes in individual tumours

Cancer is a heterogeneous disease with a strong genetic component making it suitable for precision medicine approaches aimed at identifying the underlying molecular drivers within a tumour. Large scale population-level cancer sequencing consortia have identified many actionable mutations common across both cancer types and sub-types, resulting in an increasing number of successful precision medicine programs. Nonetheless, such approaches fail to consider the effects of mutations unique to an individual patient and may miss rare driver mutations, necessitating personalised approaches to driver-gene prioritisation. One approach is to quantify the functional importance of individual mutations in a single tumour based on how they affect the expression of genes in a gene interaction network (GIN). These GIN-based approaches can be broadly divided into those that utilise an existing reference GIN and those that construct de novo patient-specific GINs. These single-tumour approaches have several limitations that likely influence their results, such as use of reference cohort data, network choice, and approaches to mathematical approximation, and more research is required to evaluate the in vitro and in vivo applicability of their predictions. This review examines the current state of the art methods that identify driver genes in single tumours with a focus on GIN-based driver prioritisation.

Quantitative Evaluation of Stem-like Markers of Human Glioblastoma Using Single-Cell RNA Sequencing Datasets

Targeting glioblastoma (GBM) stem-like cells (GSCs) is a common interest in both the laboratory investigation and clinical treatment of GBM. Most of the currently applied GBM stem-like markers lack validation and comparison with common standards regarding their efficiency and feasibility in various targeting methods. Using single-cell RNA sequencing datasets from 37 GBM patients, we obtained a large pool of 2173 GBM stem-like marker candidates. To evaluate and select these candidates quantitatively, we characterized the efficiency of the candidate markers in targeting the GBM stem-like cells by their frequencies and significance of being the stem-like cluster markers. This was followed by further selection based on either their differential expression in GBM stem-like cells compared with normal brain cells or their relative expression level compared with other expressed genes. The cellular location of the translated protein was also considered. Different combinations of selection criteria highlight different markers for different application scenarios. By comparing the commonly used GSCs marker CD133 (PROM1) with markers selected by our method regarding their universality, significance, and abundance, we revealed the limitations of CD133 as a GBM stem-like marker. Overall, we propose BCAN, PTPRZ1, SOX4, etc. for laboratory-based assays with samples free of normal cells. For in vivo targeting applications that require high efficiency in targeting the stem-like subtype, the ability to distinguish GSCs from normal brain cells, and a high expression level, we recommend the intracellular marker TUBB3 and the surface markers PTPRS and GPR56.

DrDimont: explainable drug response prediction from differential analysis of multi-omics networks

MOTIVATION

While it has been well established that drugs affect and help patients differently, personalized drug response predictions remain challenging. Solutions based on single omics measurements have been proposed, and networks provide means to incorporate molecular interactions into reasoning. However, how to integrate the wealth of information contained in multiple omics layers still poses a complex problem.

RESULTS

We present DrDimont, Drug response prediction from Differential analysis of multi-omics networks. It allows for comparative conclusions between two conditions and translates them into differential drug response predictions. DrDimont focuses on molecular interactions. It establishes condition-specific networks from correlation within an omics layer that are then reduced and combined into heterogeneous, multi-omics molecular networks. A novel semi-local, path-based integration step ensures integrative conclusions. Differential predictions are derived from comparing the condition-specific integrated networks. DrDimont's predictions are explainable, i.e. molecular differences that are the source of high differential drug scores can be retrieved. We predict differential drug response in breast cancer using transcriptomics, proteomics, phosphosite and metabolomics measurements and contrast estrogen receptor positive and receptor negative patients. DrDimont performs better than drug prediction based on differential protein expression or PageRank when evaluating it on ground truth data from cancer cell lines. We find proteomic and phosphosite layers to carry most information for distinguishing drug response.

AVAILABILITY AND IMPLEMENTATION

DrDimont is available on CRAN: https://cran.r-project.org/package=DrDimont.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Deep learning prediction of chemical-induced dose-dependent and context-specific multiplex phenotype responses and its application to personalized alzheimer’s disease drug repurposing

Predictive modeling of drug-induced gene expressions is a powerful tool for phenotype-based compound screening and drug repurposing. State-of-the-art machine learning methods use a small number of fixed cell lines as a surrogate for predicting actual expressions in a new cell type or tissue, although it is well known that drug responses depend on a cellular context. Thus, the existing approach has limitations when applied to personalized medicine, especially for many understudied diseases whose molecular profiles are dramatically different from those characterized in the training data. Besides the gene expression, dose-dependent cell viability is another important phenotype readout and is more informative than conventional summary statistics (e.g., IC50) for characterizing clinical drug efficacy and toxicity. However, few computational methods can reliably predict the dose-dependent cell viability. To address the challenges mentioned above, we designed a new deep learning model, MultiDCP, to predict cellular context-dependent gene expressions and cell viability on a specific dosage. The novelties of MultiDCP include a knowledge-driven gene expression profile transformer that enables context-specific phenotypic response predictions of novel cells or tissues, integration of multiple diverse labeled and unlabeled omics data, the joint training of the multiple prediction tasks, and a teacher-student training procedure that allows us to utilize unreliable data effectively. Comprehensive benchmark studies suggest that MultiDCP outperforms state-of-the-art methods with unseen cell lines that are dissimilar from the cell lines in the supervised training in terms of gene expressions. The predicted drug-induced gene expressions demonstrate a stronger predictive power than noisy experimental data for downstream tasks. Thus, MultiDCP is a useful tool for transcriptomics-based drug repurposing and compound screening that currently rely on noisy high-throughput experimental data. We applied MultiDCP to repurpose individualized drugs for Alzheimer’s disease in terms of efficacy and toxicity, suggesting that MultiDCP is a potentially powerful tool for personalized drug discovery.

Conventional target-based compound screening that follows the one-drug-one-gene drug discovery paradigm has a low success rate in tackling multi-genic systemic diseases such as Alzheimer’s disease. A systems pharmacology strategy is needed to target gene regulatory networks. To enable systems pharmacology-oriented phenotypic screening, it is critical to utilize a mechanistic phenotype readout to link drug responses in a model system to drug toxicity and efficacy in an individual. Chemical-induced dose-dependent gene expression profiles provide critical information on drug mode of action and off-target effects and can identify drug candidates that reverse disease phenotypes. However, state-of-the-art machine learning methods for predicting chemical-induced gene expressions are all trained using data from a limited number of cancer cell lines and can only achieve suboptimal performance when applied to new cell types or patient samples. Here, we have developed a new deep learning framework to address this challenge and demonstrated its potential in personalized drug repurposing using Alzheimer’s disease as a case study.

Application of individualized differential expression analysis in human cancer proteome

Liquid chromatography–mass spectrometry-based quantitative proteomics can measure the expression of thousands of proteins from biological samples and has been increasingly applied in cancer research. Identifying differentially expressed proteins (DEPs) between tumors and normal controls is commonly used to investigate carcinogenesis mechanisms. While differential expression analysis (DEA) at an individual level is desired to identify patient-specific molecular defects for better patient stratification, most statistical DEP analysis methods only identify deregulated proteins at the population level. To date, robust individualized DEA algorithms have been proposed for ribonucleic acid data, but their performance on proteomics data is underexplored. Herein, we performed a systematic evaluation on five individualized DEA algorithms for proteins on cancer proteomic datasets from seven cancer types. Results show that the within-sample relative expression orderings (REOs) of protein pairs in normal tissues were highly stable, providing the basis for individualized DEA for proteins using REOs. Moreover, individualized DEA algorithms achieve higher precision in detecting sample-specific deregulated proteins than population-level methods. To facilitate the utilization of individualized DEA algorithms in proteomics for prognostic biomarker discovery and personalized medicine, we provide Individualized DEP Analysis IDEPAXMBD (XMBD: Xiamen Big Data, a biomedical open software initiative in the National Institute for Data Science in Health and Medicine, Xiamen University, China.) (https://github.com/xmuyulab/IDEPA-XMBD), which is a user-friendly and open-source Python toolkit that integrates individualized DEA algorithms for DEP-associated deregulation pattern recognition.

The overexpression of DNA repair genes in invasive ductal and lobular breast carcinomas: Insights on individual variations and precision medicine

In the era of precision medicine, analyzing the transcriptomic profile of patients is essential to tailor the appropriate therapy. In this study, we explored transcriptional differences between two invasive breast cancer subtypes; infiltrating ductal carcinoma (IDC) and lobular carcinoma (LC) using RNA-Seq data deposited in the TCGA-BRCA project. We revealed 3854 differentially expressed genes between normal ductal tissues and IDC. In addition, IDC to LC comparison resulted in 663 differentially expressed genes. We then focused on DNA repair genes because of their known effects on patients' response to therapy and resistance. We here report that 36 DNA repair genes are overexpressed in a significant number of both IDC and LC patients' samples. Despite the upregulation in a significant number of samples, we observed a noticeable variation in the expression levels of the repair genes across patients of the same cancer subtype. The same trend is valid for the expression of miRNAs, where remarkable variations between patients' samples of the same cancer subtype are also observed. These individual variations could lie behind the differential response of patients to treatment. The future of cancer diagnostics and therapy will inevitably depend on high-throughput genomic and transcriptomic data analysis. However, we propose that performing analysis on individual patients rather than a big set of patients' samples will be necessary to ensure that the best treatment is determined, and therapy resistance is reduced.