Methods and approaches in the analysis of gene expression data

PurposeThis study aims to explore gene expression signatures and serum biomarkers to predict intrinsic chemoresistance in epithelial ovarian cancer (EOC).Patients and MethodsGene expression profiling data of 322 high-grade EOC cases between 2009 and 2010 in The Cancer Genome Atlas project (TCGA) were used to develop and validate gene expression signatures that could discriminate different responses to first-line platinum/paclitaxel-based treatments. A gene regulation network was then built to further identify hub genes responsible for differential gene expression between the complete response (CR) group and the progressive disease (PD) group. Further, to find more robust serum biomarkers for clinical application, we integrated our gene signatures and gene signatures reported previously to identify secretory protein-encoding genes by searching the DAVID database. In the end, gene-drug interaction network was constructed by searching Comparative Toxicogenomics Database (CTD) and literature.ResultsA 349-gene predictive model and an 18-gene model independent of key clinical features with high accuracy were developed for prediction of chemoresistance in EOC. Among them, ten important hub genes and six critical signaling pathways were identified to have important implications in chemotherapeutic response. Further, ten potential serum biomarkers were identified for predicting chemoresistance in EOC. Finally, we suggested some drugs for individualized treatment.ConclusionWe have developed the predictive models and serum biomarkers for platinum/paclitaxel response and established the new approach to discover potential serum biomarkers from gene expression profiles. The potential drugs that target hub genes are also suggested.

Analysis of gene expression data using self-organizing maps

We present a method for individual and integrative analysis of high dimension, low sample size data that capitalizes on the recurring theme in multivariate analysis of projecting higher dimensional data onto a few meaningful directions that are solutions to a generalized eigenvalue problem. We propose a general framework, called SELP (Sparse Estimation with Linear Programming), with which one can obtain a sparse estimate for a solution vector of a generalized eigenvalue problem. We demonstrate the utility of SELP on canonical correlation analysis for an integrative analysis of methylation and gene expression profiles from a breast cancer study, and we identify some genes known to be associated with breast carcinogenesis, which indicates that the proposed method is capable of generating biologically meaningful insights. Simulation studies suggest that the proposed method performs competitive in comparison with some existing methods in identifying true signals in various underlying covariance structures.

Background - In the NeoTRIPaPDL1 phase III trial, triple-negative breast cancer (TNBC) patients were randomized to receive nab-paclitaxel/carboplatin for 8 cycles (CT arm) with or without atezolizumab (CT/A arm). We previously reported that the majority of patients (65%) with PD-L1- at baseline converted to PD-L1+ (the majority with IC2/IC3) after the first cycle of treatment in the atezolizumab arm. Here we studied treatment associated changes in the molecular tumour features and immune microenvironment by PD-L1 expression. Methods – A total of 158 (56.4%) of patients enrolled in the NeoTRIPaPDL1 trial (76 and 82 from CT/A and CT arm, respectively) were included in this analysis by satisfying the following criteria: i) availability of a paired core biopsy at baseline and at day 1 of the second treatment cycle (D1C2); ii) evaluation of stromal TILs (sTILs) and staining for PD-L1 (Ventana SP142); iii) successful RNA-seq gene expression profiling. PD-L1 groups were defined as IC0/IC1 (<5% PD-L1low) vs IC2/IC3 (>=5% PD-L1high). Presence of different immune cell populations was quantified by gene expression profile deconvolution using ConsensusTME R package. Cancer hallmark gene set collection and custom signature activation status were estimated in each sample using singscore R package. Differences in score distribution were evaluated by 2-sided t-test. Results At baseline, both sTILs and immune cell signatures were upregulated in PD-L1high compared to PD-L1low tumours (p<0.001). No significant differences were found between the two treatment arms in the PD-L1high subpopulation. In the PD-L1low cases, both sTILs and immune related signatures were slightly downregulated in the CT/A arm (p<0.05), suggesting a modest unbalance among treatment arms with CT arm being slightly more inflamed than CT/A arm.At D1C2, PD-L1high tumours in the CT arm systematically had high sTILs (median=70%, range=30-90%), while PD-L1low tumours receiving CT/A had low sTILs in a significant proportion of cases (median=30%, range=0-90%; p<0.001). Similarly, at D1C2 several gene expression-estimated immune cell populations and immune-related signatures were upregulated in the CT arm compared to CT/A arm in PD-L1high tumours, with the weakest association observed for M2 macrophages (p=0.058). No tumour-related signatures were differentially expressed among the two treatment arms within groups with PD-L1high, suggesting that different treatment. modulate PD-L1 by engaging a different immune mileau instead of modulating tumor related features.Considering PD-L1low groups at D1C2, in the CT arm 21 tumors had >30% sTILs (n=21/69, 30.4%), while in the CT/A arm only 2 had sTILs >30% (2/27, 7.4%) (p<0.001). Analysis of gene expression data identified IFN-related signatures as the most upregulated in the CT compared to CT/A arm in PD-L1low cases at D1C2 (p<0.05). Conclusions Integrated dynamic analysis of PD-L1 expression and gene expression data highlighted significant treatment-specific changes of the immune landscape according to PD-L1 expression, when this biomarker is assessed during treatment. This indicates that the immune milieu associated with PD-L1 status is strongly dependent on the time of assessment in relationship to treatment received. Such observation may explain why in the NeoTRIP trial, baseline PD-L1 but not on-treatment PD-L1 was predictive of pCR in CT/A arm. In addition, our findings could have implications related to the use of PD-L1 as a predictive biomarker in pre-treated patients, especially when assessed early on during treatment. Citation Format: Maurizio Callari, Chiun-Sheng Huang, Daniel Egle, Begoña Bermejo, Claudio Zamagni, Matteo Dugo, Marc Thill, Antonio Anton, Marco Barreca, Stefania Russo, Eva Maria Ciruelos, Richard Greil, Stefania Zambelli, Balázs Gyorffy, Chanel Smart, Olivia Biasi, Pinuccia Valagussa, Giuseppe Viale, Luca Gianni, Giampaolo Bianchini. Immune milieu associated with PD-L1 status in TNBC is dependent on time of biomarker assessment and treatment received: A secondary analysis of the NeoTRIPaPDL1 trial [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr P1-04-02.

Background. Precisely predicting cancer is crucial for cancer treatment. Gene expression profiles make it possible to analyze patterns between genes and cancers on the genome-wide scale. Gene expression data analysis, however, is confronted with enormous challenges for its characteristics, such as high dimensionality, small sample size, and low Signal-to-Noise Ratio. Results. This paper proposes a method, termed RS_SVM, to predict gene expression profiles via aggregating SVM trained on random subspaces. After choosing gene features through statistical analysis, RS_SVM randomly selects feature subsets to yield random subspaces and training SVM classifiers accordingly and then aggregates SVM classifiers to capture the advantage of ensemble learning. Experiments on eight real gene expression datasets are performed to validate the RS_SVM method. Experimental results show that RS_SVM achieved better classification accuracy and generalization performance in contrast with single SVM, K-nearest neighbor, decision tree, Bagging, AdaBoost, and the state-of-the-art methods. Experiments also explored the effect of subspace size on prediction performance. Conclusions. The proposed RS_SVM method yielded superior performance in analyzing gene expression profiles, which demonstrates that RS_SVM provides a good channel for such biological data.

8503 Background: Small pilot studies (e.g., N Engl J Med. 2018;378:1976) have shown that preoperative immune checkpoint inhibitor therapy may be of benefit in early-stage NSCLC. This large multicenter trial assesses the benefit of neoadjuvant treatment with atezolizumab (atezo; NCT02927301). Methods: Patients (pts) with stages IB to selected IIIB resectable NSCLC receive 2 cycles of atezo 1200 mg (days 1, 22) then undergo resection (day 40 ± 10). Primary tumor +/- node biopsies and blood samples are obtained before atezo and at surgery for biomarker studies. The primary endpoint is major pathological response (MPR), defined as ≤ 10% viable tumor cells in the resection specimen. Secondary endpoints include safety and correlation of response with PD-L1 expression, tumor mutation burden (TMB) and gene expression signatures. Results: For this interim efficacy analysis (5 Sep 2018 data cut), we report on the first 101 of 180 planned pts: 47 males, median age, 64 y; all ECOG PS 0-1; 23 current and 68 former smokers; 66 non-squamous NSCLC; clinical stages IB/IIA/IIB/IIIA/IIIB n = 11/16/28/39/7. There were 2 treatment-unrelated Gr 5 AEs (cardiac death post surgical resection; death due to disease progression), 29 Gr 3-4 AEs (6 [6%] treatment related). 90 pts had surgery. Excluding 8 pts who had driver mutations (7 EGFR, 1 ALK, no MPR), MPR rate was 15/82 (18%, 95% CI 11%-28%), 4 pts had pathological complete response (pCR). By RECIST, 6/82 pts had PR, 72 had SD and 4 had PD. Two of 26 (8%) PD-L1− (TC0 and IC0, clone SP142) and 10 of 35 (29%) PD-L1+ had MPR ( P= 0.055). Five of 44 (11%) TPS < 50 (PD-L1 clone 22C3) and 7 of 20 (35%) TPS > 50 had MPR ( P= 0.040). Exome sequencing data was available for 47/101 pts. Median TMB was 10.4 (range, 1.5-46.5) mutations per Mb and was not different in those with MPR compared with those without MPR. Further analysis of TMB, mutation signatures, and gene expression profiling is ongoing. Conclusions: Atezo in the neoadjuvant setting was well tolerated, and pCR and MPR rates are encouraging in this large multicenter trial. Efficacy interim analysis passed its futility boundary, and study enrollment continues. Safety, efficacy results and ongoing correlative analyses will be presented. Clinical trial information: NCT02927301.

BackgroundSerial Analysis of Gene Expression (SAGE) is a DNA sequencing-based method for large-scale gene expression profiling that provides an alternative to microarray analysis. Most analyses of SAGE data aimed at identifying co-expressed genes have been accomplished using various versions of clustering approaches that often result in a number of false positives.Principal FindingsHere we explore the use of seriation, a statistical approach for ordering sets of objects based on their similarity, for large-scale expression pattern discovery in SAGE data. For this specific task we implement a seriation heuristic we term ‘progressive construction of contigs’ that constructs local chains of related elements by sequentially rearranging margins of the correlation matrix. We apply the heuristic to the analysis of simulated and experimental SAGE data and compare our results to those obtained with a clustering algorithm developed specifically for SAGE data. We show using simulations that the performance of seriation compares favorably to that of the clustering algorithm on noisy SAGE data.ConclusionsWe explore the use of a seriation approach for visualization-based pattern discovery in SAGE data. Using both simulations and experimental data, we demonstrate that seriation is able to identify groups of co-expressed genes more accurately than a clustering algorithm developed specifically for SAGE data. Our results suggest that seriation is a useful method for the analysis of gene expression data whose applicability should be further pursued.

Integrated analysis of global gene and microRNA expression profiling associated with aplastic anaemia

The recent discovery of ‘molecular subtypes’ in human primary colorectal cancer has revealed correlations between subtype, propensity to metastasize and response to therapy. It is currently not known whether the molecular tumor subtype is maintained after distant spread. If this is the case, molecular subtyping of the primary tumor could guide subtype-targeted therapy of metastatic disease. In this study, we classified paired samples of primary colorectal carcinomas and their corresponding liver metastases (n=129) as epithelial-like or mesenchymal-like, using a recently developed immunohistochemistry-based classification tool. We observed considerable discordance (45%) in the classification of primary tumors and their liver metastases. Discordant classification was significantly associated with the use of neoadjuvant chemotherapy. Furthermore, gene expression analysis of chemotherapy-exposed versus chemotherapy naive liver metastases revealed expression of a mesenchymal program in pre-treated tumors. To explore whether chemotherapy could cause gene expression changes influencing molecular subtyping, we exposed patient-derived colonospheres to six short cycles of 5-fluorouracil. Gene expression profiling and signature enrichment analysis subsequently revealed that the expression of signatures identifying mesenchymal-like tumors was strongly increased in chemotherapy-exposed tumor cultures. Unsupervised clustering of large cohorts of human colon tumors with the chemotherapy-induced gene expression program identified a poor prognosis mesenchymal-like subgroup. We conclude that neoadjuvant chemotherapy induces a mesenchymal phenotype in residual tumor cells and that this may influence the molecular classification of colorectal tumors.

The application of global gene expression profiling allows to obtain detailed molecular fingerprints of underlying gene expression in any cell of interest. In this work gene expression profiles were generated from a comprehensive cohort of leukemia patients and healthy donors referred to and diagnosed in the Laboratory for Leukemia Diagnostics, Munich, Germany, which is a nation-wide reference center for the diagnosis of hematologic malignancies. Thoroughly characterized clinical samples were analyzed by high-density microarrays interrogating the expression status of more than 33,000 transcripts. In one specific aspect of this work the potential application of gene expression signatures for the prediction and classification of specific leukemia subtypes was assessed. Today the diagnosis and subclassification of leukemias is based on a controlled application of various techniques including cytomorphology, cytogenetics, fluorescence in situ hybridization, multiparameter flow cytometry, and PCR-based methods. The diagnostic procedure is performed according to a specific algorithm, but is time-consuming, cost-intensive, and requires expert knowledge. Based on a very low number of candidate genes it is demonstrated in this work that prognostically relevant acute leukemia subtypes can be classified using microarray technology. Moreover, in an expanded analysis including 937 patient samples representing 12 distinct clinically relevant acute and chronic leukemia subtypes and healthy, non-leukemia bone marrow specimens a diagnostic prediction accuracy of ~95% was achieved. Thus, given these results it can be postulated that the occurring patterns in gene expression would be so robust that they would allow to predict the leukemia subtype using global gene expression profiling technology. This finding is further substantiated through the demonstration that reported differentially expressed genes from the literature, namely pediatric gene expression signatures representing various acute lymphoblastic leukemia (ALL) subtypes, can be used to independently predict the corresponding adult ALL subtypes. Furthermore, it could be demonstrated that microarrays both confirm and reproduce data from standard diagnostic procedures, but also provide very robust results. Parameters such as partial RNA degradation, shipment time of the samples, varying periods of storage of the samples, or target preparations at different time points from either bone marrow or peripheral blood specimens by different operators did not dramatically influence the diagnostic gene expression signatures. In another major aspect of this work gene expression signatures were examined in detail to obtain new insights into the underlying biology of acute promyelocytic leukemia (APL) and t(11q23)/MLL leukemias. In APL, microarrays led to a deeper understanding of morphological and clinical characteristics. Firstly, genes which have a functional relevance in blood coagulation were found to be differentially expressed when APL was compared to other acute myeloid leukemia (AML) subtypes. Secondly, a supervised pairwise comparison between the two different APL phenotypes, M3 and its variant M3v, for the first time revealed differentially expressed genes encoding for biological functions and pathways such as granulation and maturation. With respect to 11q23 leukemias it could be demonstrated that leukemias with rearrangements of the MLL gene are characterized by a common specific gene expression signature. Additionally, in unsupervised and supervised data analysis algorithms ALL and AML cases with t(11q23)/MLL segregated according to the lineage, i.e., myeloid or lymphoid, respectively. This segregation could be explained by a highly differing transcriptional program. Through the use of biological network analyses essential regulators of early B cell development, PAX5 and EBF, were shown to be associated with a clear B-lineage commitment in lymphoblastic t(11q23)/MLL leukemias. Also, the influence of the different MLL translocation partners on the transcriptional program was directly assessed. But interestingly, gene expression profiles did not reveal a clear distinct pattern associated with one of the analyzed partner genes. Taken together, the identified molecular expression pattern of MLL fusion gene samples and biological networks revealed new insights into the aberrant transcriptional program in t(11q23)/MLL leukemias. In addition, a series of analyses was targeted to obtain new insights into the underlying biology in heterogeneous B-lineage leukemias not positive for BCR/ABL or MLL gene rearrangements. It could be demonstrated that the genetically more heterogeneous precursor B-ALL samples intercalate with BCR/ABL-positive cases, but their profiles were clearly distinct from T-ALL and t(11q23)/MLL cases. In conclusion, various unsupervised and supervised data analysis strategies demonstrated that defined leukemia subtypes can be characterized on the basis of distinct gene expression signatures. Specific gene expression patterns reproduced the taxonomy of this hematologic malignancy, provided new insights into different disease subtypes, and identified critical pathway components that might be considered for future therapeutic intervention. Based on these results it is now further possible to develop a one-step diagnostic approach for the diagnosis of leukemias using a customized microarray.

The problems of analyzing dose effects on gene expression are gaining attention in toxicological research. Determining how gene expression profiles change with toxicant dose will improve the utility of arrays in identifying biomarkers and elucidating their modes of toxic action. In the present study, we focused on determining the dose-dependent alterations of gene expression profiles with hexanal exposure and we identified the possible biomarkers of hexanal in A549 human alveolar cells. A549 cells were exposed to a 5% inhibitory concentration (IC5) and a 20% inhibitory concentration (IC20) of hexanal for 48 h. Through microarray analysis using an oligonucleotide chip, we identified that the gene expression patterns were differentially shown in the control group and the hexanalexposed groups. The hexanal-exposed groups are more sensitive to gene alteration than the control group, and gene expressions are more significantly altered in the IC20 exposure group than in the IC5 exposure group. With clustering analysis of gene expression profiles, we identified 2,929 IC5− and 3,678 IC20-specific genes, and 302 dose-dependently expressed genes. Gene ontology (GO) analysis with 246 annotated genes of the 302 dosedependent expressed genes showed correlation with the key biological processes involved in neurological system processes, immune system development, cell activation, and cell-cell signaling. In conclusion, current study describes alterations in gene expression profiles in response to exposure to different doses of hexanal and related toxic pathways induced by significantly expressed genes. Moreover, novel genes and pathways that could potentially play a role in the prevention of respiratory disease due to aldehydes are identified.

Genes that are highly overexpressed in tumor cells can be required for tumor cell survival and have the potential to be selective therapeutic targets. In an attempt to identify such targets, we combined a functional genomics and a systems biology approach to assess the consequences of RNAi-mediated silencing of overexpressed genes that were selected from 140 gene expression profiles from colorectal cancers (CRCs) and matched normal mucosa. In order to identify credible models for in-depth functional analysis, we first confirmed the overexpression of these genes in 25 different CRC cell lines. We then identified five candidate genes that profoundly reduced the viability of CRC cell lines when silenced with either siRNAs or short-hairpin RNAs (shRNAs), i.e., HMGA1, TACSTD2, RRM2, RPS2 and NOL5A. These genes were further studied by systematic analysis of comprehensive gene expression profiles generated following siRNA-mediated silencing. Exploration of these RNAi-specific gene expression signatures allowed the identification of the functional space in which the five genes operate and showed enrichment for cancer-specific signaling pathways, some known to be involved in CRC. By comparing the expression of the RNAi signature genes with their respective expression levels in an independent set of primary rectal carcinomas, we could recapitulate these defined RNAi signatures, therefore, establishing the biological relevance of our observations. This strategy identified the signaling pathways that are affected by the prominent oncogenes HMGA1 and TACSTD2, established a yet unknown link between RRM2 and PLK1 and identified RPS2 and NOL5A as promising potential therapeutic targets in CRC.

Analysis of gene expression profiles under water stress in tolerant and sensitive sugarcane plants

Computational analysis methods including machine learning have a significant impact in the fields of genomics and medicine. High-throughput gene expression analysis methods such as microarray technology and RNA sequencing produce enormous amounts of data. Traditionally, statistical methods are used for comparative analysis of gene expression data. However, more complex analysis for classification of sample observations, or discovery of feature genes requires sophisticated computational approaches. In this review, we compile various statistical and computational tools used in analysis of expression microarray data. Even though the methods are discussed in the context of expression microarrays, they can also be applied for the analysis of RNA sequencing and quantitative proteomics datasets. We discuss the types of missing values, and the methods and approaches usually employed in their imputation. We also discuss methods of data normalization, feature selection, and feature extraction. Lastly, methods of classification and class discovery along with their evaluation parameters are described in detail. We believe that this detailed review will help the users to select appropriate methods for preprocessing and analysis of their data based on the expected outcome.

Protein kinase C (PKC) isozymes are key signal transducers involved in normal physiology and disease and have been widely implicated in cancer progression. Despite our extensive knowledge of the signaling pathways regulated by PKC isozymes and their effectors, there is essentially no information on how individual members of the PKC family regulate gene transcription. Here, we report the first PKC isozyme-specific analysis of global gene expression by microarray using RNAi depletion of diacylglycerol/phorbol ester-regulated PKCs. A thorough analysis of this microarray data revealed unique patterns of gene expression controlled by PKCα, PKCδ, and PKCε, which are remarkably different in cells growing in serum or in response to phorbol ester stimulation. PKCδ is the most relevant isoform in controlling the induction of genes by phorbol ester stimulation, whereas PKCε predominantly regulates gene expression in serum. We also established that two PKCδ-regulated genes, FOSL1 and BCL2A1, mediate the apoptotic effect of phorbol esters or the chemotherapeutic agent etoposide in prostate cancer cells. Our studies offer a unique opportunity for establishing novel transcriptional effectors for PKC isozymes and may have significant functional and therapeutic implications.