Batch Effect Correction Research Articles

Identification of transcriptional programs using dense vector representations defined by mutual information with GeneVector

Deciphering individual cell phenotypes from cell-specific transcriptional processes requires high dimensional single cell RNA sequencing. However, current dimensionality reduction methods aggregate sparse gene information across cells, without directly measuring the relationships that exist between genes. By performing dimensionality reduction with respect to gene co-expression, low-dimensional features can model these gene-specific relationships and leverage shared signal to overcome sparsity. We describe GeneVector, a scalable framework for dimensionality reduction implemented as a vector space model using mutual information between gene expression. Unlike other methods, including principal component analysis and variational autoencoders, GeneVector uses latent space arithmetic in a lower dimensional gene embedding to identify transcriptional programs and classify cell types. In this work, we show in four single cell RNA-seq datasets that GeneVector was able to capture phenotype-specific pathways, perform batch effect correction, interactively annotate cell types, and identify pathway variation with treatment over time.

Single-cell RNA sequencing deciphers transcriptional profiles of hepatocytes in mouse with hepatic alveolar echinococcosis

To investigate the cell composition and the transcriptional characteristics in microenvironments of hepatic tissues in mice at late stage of Echinococcus multilocularis infection at a single-cell level. Peri-lesion and paired distal hepatic specimens were collected from two BALB/c mice (6 to 8 weeks old) infected with E. multilocularis for single-cell RNA sequencing. The Seurat package in the R software was employed for quality control of data, multi-sample integration and correction of batch effects, and uniform manifold approximation and projection (UMAP) algorithm was used for cell clustering. Cell types were annotated using classical marker genes. Differentially expressed genes were screened in each cell type through differential gene expression analysis, and the biological roles of cells were predicted using Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses. A total of 43 710 cells from peri-lesion and distal hepatic tissues of E. multilocularis-infected mice were analyzed, and were classified into 11 cell types, including neutrophils, T cells, macrophages, granulocyte-monocyte progenitor cells, B cells, plasma cells, basophils, hepatic stellate cells, endothelial cells, hepatocytes, and platelets. T cells were the largest population of immune cells in the microenvironment of hepatic tissues, including five CD4+ T cell subsets, two CD8+ T cell subsets and phosphoantigen-reactive γδT cells. The proportions of CD4+ helper T cells and cytotoxic CD4+ T cells decreased and the proportion of T helper 2 (Th2) cells increased in peri-lesion tissues relative to distal hepatic tissues. In addition, the differentially expressed genes in Th2 cells were associated with negative regulation of the immune system, and the highly expressed genes in cytotoxic CD4+ T cells correlated with activation of the immune system. Single-cell RNA sequencing deciphers the cell composition and distribution in microenvironments of hepatic tissues from mice infected with E. multilocularis, and the increased proportion of Th2 cells in peri-lesion hepatic tissues may be associated with formation of immunosuppressive microenvironments.

Innovative breakthroughs facilitated by single-cell multi-omics: manipulating natural killer cell functionality correlates with a novel subcategory of melanoma cells.

Melanoma is typically regarded as the most dangerous form of skin cancer. Although surgical removal of in situ lesions can be used to effectively treat metastatic disease, this condition is still difficult to cure. Melanoma cells are removed in great part due to the action of natural killer (NK) and T cells on the immune system. Still, not much is known about how the activity of NK cell-related pathways changes in melanoma tissue. Thus, we performed a single-cell multi-omics analysis on human melanoma cells in this study to explore the modulation of NK cell activity. Cells in which mitochondrial genes comprised > 20% of the total number of expressed genes were removed. Gene ontology (GO), gene set enrichment analysis (GSEA), gene set variation analysis (GSVA), and AUCcell analysis of differentially expressed genes (DEGs) in melanoma subtypes were performed. The CellChat package was used to predict cell-cell contact between NK cell and melanoma cell subtypes. Monocle program analyzed the pseudotime trajectories of melanoma cells. In addition, CytoTRACE was used to determine the recommended time order of melanoma cells. InferCNV was utilized to calculate the CNV level of melanoma cell subtypes. Python package pySCENIC was used to assess the enrichment of transcription factors and the activity of regulons in melanoma cell subtypes. Furthermore, the cell function experiment was used to confirm the function of TBX21 in both A375 and WM-115 melanoma cell lines. Following batch effect correction, 26,161 cells were separated into 28 clusters and designated as melanoma cells, neural cells, fibroblasts, endothelial cells, NK cells, CD4+ T cells, CD8+ T cells, B cells, plasma cells, monocytes and macrophages, and dendritic cells. A total of 10137 melanoma cells were further grouped into seven subtypes, i.e., C0 Melanoma BIRC7, C1 Melanoma CDH19, C2 Melanoma EDNRB, C3 Melanoma BIRC5, C4 Melanoma CORO1A, C5 Melanoma MAGEA4, and C6 Melanoma GJB2. The results of AUCell, GSEA, and GSVA suggested that C4 Melanoma CORO1A may be more sensitive to NK and T cells through positive regulation of NK and T cell-mediated immunity, while other subtypes of melanoma may be more resistant to NK cells. This suggests that the intratumor heterogeneity (ITH) of melanoma-induced activity and the difference in NK cell-mediated cytotoxicity may have caused NK cell defects. Transcription factor enrichment analysis indicated that TBX21 was the most important TF in C4 Melanoma CORO1A and was also associated with M1 modules. In vitro experiments further showed that TBX21 knockdown dramatically decreases melanoma cells' proliferation, invasion, and migration. The differences in NK and T cell-mediated immunity and cytotoxicity between C4 Melanoma CORO1A and other melanoma cell subtypes may offer a new perspective on the ITH of melanoma-induced metastatic activity. In addition, the protective factors of skin melanoma, STAT1, IRF1, and FLI1, may modulate melanoma cell responses to NK or T cells.

How missing value imputation is confounded with batch effects and what you can do about it

In data-processing pipelines, upstream steps can influence downstream processes because of their sequential nature. Among these data-processing steps, batch effect (BE) correction (BEC) and missing value imputation (MVI) are crucial for ensuring data suitability for advanced modeling and reducing the likelihood of false discoveries. Although BEC-MVI interactions are not well studied, they are ultimately interdependent. Batch sensitization can improve the quality of MVI. Conversely, accounting for missingness also improves proper BE estimation in BEC. Here, we discuss how BEC and MVI are interconnected and interdependent. We show how batch sensitization can improve any MVI and bring attention to the idea of BE-associated missing values (BEAMs). Finally, we discuss how batch-class imbalance problems can be mitigated by borrowing ideas from machine learning.

Pooled analysis of frontal lobe transcriptomic data identifies key mitophagy gene changes in Alzheimer's disease brain.

The growing prevalence of Alzheimer's disease (AD) is becoming a global health challenge without effective treatments. Defective mitochondrial function and mitophagy have recently been suggested as etiological factors in AD, in association with abnormalities in components of the autophagic machinery like lysosomes and phagosomes. Several large transcriptomic studies have been performed on different brain regions from AD and healthy patients, and their data represent a vast source of important information that can be utilized to understand this condition. However, large integration analyses of these publicly available data, such as AD RNA-Seq data, are still missing. In addition, large-scale focused analysis on mitophagy, which seems to be relevant for the aetiology of the disease, has not yet been performed. In this study, publicly available raw RNA-Seq data generated from healthy control and sporadic AD post-mortem human samples of the brain frontal lobe were collected and integrated. Sex-specific differential expression analysis was performed on the combined data set after batch effect correction. From the resulting set of differentially expressed genes, candidate mitophagy-related genes were identified based on their known functional roles in mitophagy, the lysosome, or the phagosome, followed by Protein-Protein Interaction (PPI) and microRNA-mRNA network analysis. The expression changes of candidate genes were further validated in human skin fibroblast and induced pluripotent stem cells (iPSCs)-derived cortical neurons from AD patients and matching healthy controls. From a large dataset (AD: 589; control: 246) based on three different datasets (i.e., ROSMAP, MSBB, & GSE110731), we identified 299 candidate mitophagy-related differentially expressed genes (DEG) in sporadic AD patients (male: 195, female: 188). Among these, the AAA ATPase VCP, the GTPase ARF1, the autophagic vesicle forming protein GABARAPL1 and the cytoskeleton protein actin beta ACTB were selected based on network degrees and existing literature. Changes in their expression were further validated in AD-relevant human in vitro models, which confirmed their down-regulation in AD conditions. Through the joint analysis of multiple publicly available data sets, we identify four differentially expressed key mitophagy-related genes potentially relevant for the pathogenesis of sporadic AD. Changes in expression of these four genes were validated using two AD-relevant human in vitro models, primary human fibroblasts and iPSC-derived neurons. Our results provide foundation for further investigation of these genes as potential biomarkers or disease-modifying pharmacological targets.

Batch effect correction methods for NASA GeneLab transcriptomic datasets

Introduction:RNA sequencing (RNA-seq) data from space biology experiments promise to yield invaluable insights into the effects of spaceflight on terrestrial biology. However, sample numbers from each study are low due to limited crew availability, hardware, and space. To increase statistical power, spaceflight RNA-seq datasets from different missions are often aggregated together. However, this can introduce technical variation or “batch effects”, often due to differences in sample handling, sample processing, and sequencing platforms. Several computational methods have been developed to correct for technical batch effects, thereby reducing their impact on true biological signals.Methods:In this study, we combined 7 mouse liver RNA-seq datasets from NASA GeneLab (part of the NASA Open Science Data Repository) to evaluate several common batch effect correction methods (ComBat and ComBat-seq from the sva R package, and Median Polish, Empirical Bayes, and ANOVA from the MBatch R package). Principal component analysis (PCA) was used to identify library preparation method and mission as the primary sources of batch effect among the technical variables in the combined dataset. We next quantitatively evaluated the ability of each of the indicated methods to correct for each identified technical batch variable using the following criteria: BatchQC, PCA, dispersion separability criterion, log fold change correlation, and differential gene expression analysis. Each batch variable/correction method combination was then assessed using a custom scoring approach to identify the optimal correction method for the combined dataset, by geometrically probing the space of all allowable scoring functions to yield an aggregate volume-based scoring measure.Results and Discussion:Using the method described for the combined dataset in this study, the library preparation variable/ComBat correction method pair out ranked the other candidate pairs, suggesting that this combined dataset should be corrected for library preparation using the ComBat correction method prior to downstream analysis. We describe the GeneLab multi-study analysis and visualization portal which will allow users to access the publicly available space biology ‘omics data, select multiple studies to combine for analysis, and examine the presence or absence of batch effects using multiple metrics. If the user chooses to perform batch effect correction, the scoring approach described here can be implemented to identify the optimal correction method to use for their specific combined dataset prior to analysis.

Advanced deep-learning algorithm for multi-cancer detection using cf-WGS.

e13560 Background: Applying machine learning in circulating cell-free DNA (cfDNA) whole genome sequencing (WGS) holds the potential to be a valuable tool for early cancer detection. However, unless the batch effect is taken into account, detecting very tiny cancer signals involves an overfitting issue with relatively small training sample number for artificial intelligence training. This study aimed to develop advanced deep-learning model for cancer detection from cfDNA whole genome sequencing data with minimal batch effect. Methods: We generated low depth cfDNA whole genome sequencing data from 412 cancer patients and 1,269 healthy individuals. The size and end motif frequency of each fragment was measured and represented as a two-dimensional matrix. Each sample data was normalized with reference samples in the same batch prior to model training to remove batch effects between sequencing runs and was divided into 4 datasets stratified by sequencing batch for cross-validation of the models. The final model combines a grouped convolution neural network (CNN) and a long short-term memory (LSTM) model for sequential size information training. Results: The model showed an accuracy of 93.1% and an AUC of 0.97. With a 95% specificity threshold, the model showed an overall sensitivity of 87% and a precision of 84.7%. For individual cancer types, liver, ovarian, esophageal, pancreatic, and lung showed sensitivities of 94.4%, 81.1%, 85.6%, 100%, and 78.2%, respectively. For cancer stages, the sensitivity was 78.6% in stage I, 86.5% in stage II, 91.5% in stage III, and 92.7% in stage IV at 95% specificity threshold. Conclusions: The deep learning models trained using fragment size and end motif frequency of cfDNA demonstrated promising results for cancer detection. The improved performance of this model highlights the potential for improving cancer detection by incorporating advanced deep learning algorithm with well curated training data for batch effect correction.

Instrumental Drift in Untargeted Metabolomics: Optimizing Data Quality with Intrastudy QC Samples.

Untargeted metabolomics is an important tool in studying health and disease and is employed in fields such as biomarker discovery and drug development, as well as precision medicine. Although significant technical advances were made in the field of mass-spectrometry driven metabolomics, instrumental drifts, such as fluctuations in retention time and signal intensity, remain a challenge, particularly in large untargeted metabolomics studies. Therefore, it is crucial to consider these variations during data processing to ensure high-quality data. Here, we will provide recommendations for an optimal data processing workflow using intrastudy quality control (QC) samples that identifies errors resulting from instrumental drifts, such as shifts in retention time and metabolite intensities. Furthermore, we provide an in-depth comparison of the performance of three popular batch-effect correction methods of different complexity. By using different evaluation metrics based on QC samples and a machine learning approach based on biological samples, the performance of the batch-effect correction methods were evaluated. Here, the method TIGER demonstrated the overall best performance by reducing the relative standard deviation of the QCs and dispersion-ratio the most, as well as demonstrating the highest area under the receiver operating characteristic with three different probabilistic classifiers (Logistic regression, Random Forest, and Support Vector Machine). In summary, our recommendations will help to generate high-quality data that are suitable for further downstream processing, leading to more accurate and meaningful insights into the underlying biological processes.

ScRAA: the development of a robust and automatic annotation procedure for single-cell RNA sequencing data

ABSTRACT A critical task in single-cell RNA sequencing (scRNA-Seq) data analysis is to identify cell types from heterogeneous tissues. While the majority of classification methods demonstrated high performance in scRNA-Seq annotation problems, a robust and accurate solution is desired to generate reliable outcomes for downstream analyses, for instance, marker genes identification, differentially expressed genes, and pathway analysis. It is hard to establish a universally good metric. Thus, a universally good classification method for all kinds of scenarios does not exist. In addition, reference and query data in cell classification are usually from different experimental batches, and failure to consider batch effects may result in misleading conclusions. To overcome this bottleneck, we propose a robust ensemble approach to classify cells and utilize a batch correction method between reference and query data. We simulated four scenarios that comprise simple to complex batch effect and account for varying cell-type proportions. We further tested our approach on both lung and pancreas data. We found improved prediction accuracy and robust performance across simulation scenarios and real data. The incorporation of batch effect correction between reference and query, and the ensemble approach improve cell-type prediction accuracy while maintaining robustness. We demonstrated these through simulated and real scRNA-Seq data.

MBECS: Microbiome Batch Effects Correction Suite

Despite the availability of batch effect correcting algorithms (BECA), no comprehensive tool that combines batch correction and evaluation of the results exists for microbiome datasets. This work outlines the Microbiome Batch Effects Correction Suite development that integrates several BECAs and evaluation metrics into a software package for the statistical computation framework R.

Concordance-Based Batch Effect Correction for Large-Scale Metabolomics.

For a large-scale metabolomics study, sample collection, preparation, and analysis may last several days, months, or even (intermittently) over years. This may lead to apparent batch effects in the acquired metabolomics data due to variability in instrument status, environmental conditions, or experimental operators. Batch effects may confound the true biological relationships among metabolites and thus obscure real metabolic changes. At present, most of the commonly used batch effect correction (BEC) methods are based on quality control (QC) samples, which require sufficient and stable QC samples. However, the quality of the QC samples may deteriorate if the experiment lasts for a long time. Alternatively, isotope-labeled internal standards have been used, but they generally do not provide good coverage of the metabolome. On the other hand, BEC can also be conducted through a data-driven method, in which no QC sample is needed. Here, we propose a novel data-driven BEC method, namely, CordBat, to achieve concordance between each batch of samples. In the proposed CordBat method, a reference batch is first selected from all batches of data, and the remaining batches are referred to as "other batches." The reference batch serves as the baseline for the batch adjustment by providing a coordinate of correlation between metabolites. Next, a Gaussian graphical model is built on the combined dataset of reference and other batches, and finally, BEC is achieved by optimizing the correction coefficients in the other batches so that the correlation between metabolites of each batch and their combinations are in concordance with that of the reference batch. Three real-world metabolomics datasets are used to evaluate the performance of CordBat by comparing it with five commonly used BEC methods. The present experimental results showed the effectiveness of CordBat in batch effect removal and the concordance of correlation between metabolites after BEC. CordBat was found to be comparable to the QC-based methods and achieved better performance in the preservation of biological effects. The proposed CordBat method may serve as an alternative BEC method for large-scale metabolomics that lack proper QC samples.

Abstract LB079: An end-to-end Visium spatial transcriptomics computational pipeline for generating low-code interactive reports of spatial insights

Abstract In recent years, methods for analyzing and interpreting expression data have evolved rapidly. Newer and more complex datasets are being developed faster than ever, and the resources needed to perform rigorous data analysis have increased dramatically. In spatial transcriptomics analysis, researchers face many technical challenges: data management, compute power, and programming knowledge. In addition to these technical barriers, a constantly evolving field can make it difficult for scientists to understand if they are performing adequate quality control, proper data integration, choosing situationally correct methods, and showing accurate biological insights. Transcriptomics data comes with many meaningful caveats that make having the proper technical resources just a single piece of the puzzle. These obstacles keep scientists with valuable biological questions from maximizing the power of Spatial Transcriptomics. To address these challenges, we developed an end-to-end Visium spatial transcriptomics computational pipeline for generating interactive figures and reports for low-code exploration. This pipeline automated data quality control, batch-effects correction, and multiple peer-reviewed analysis techniques to perform a spatially resolved analysis of tissue heterogeneity. To demonstrate the effectiveness of this pipeline, we performed an analysis on a cohort consisting of 3 patients with adenocarcinoma and 1 patient with signet ring cell carcinoma, each with 1 sample of primary colon tumor and its matched adjacent normal tissue. All samples were formalin-fixed paraffin-embedded (FFPE) provided by BioChain Institute. The tissue sections were stained with Hematoxylin and Eosin, and the transcriptome was mapped using 10x Genomics Visium Spatial Gene Expression for FFPE. The raw base call files were then processed into feature-barcode matrices, followed by quality control measures that included data filtering, sample integration, batch-effect removal, and normalization. The pipeline then combined multiple analysis methods for feature extraction from genes and spatial spots, including clustering, neighborhood detection, and gene module detection. Further, we leveraged clinical metadata to perform differential expression and abundance analyses of the extracted features across different tissue and disease types. The pipeline finally generated an interactive report detailing analysis of spatially-resolved tissue biology, pathology, and microenvironment, allowing a scientist to analyze and explore the data further. This pipeline to identified differences in the spatial arrangement between tumor and normal tissue, in addition to demonstrating the spatial arrangements that differentiate adenocarcinoma and signet ring cell carcinoma relating to the invasive front and body of the tumors. Citation Format: Steven Hamel, Elim Cheung, Ying Qu, Malachi Loviska, Aaron Mayer, Lutong Zhang, Tong Lu, Vidyodhaya Sundaram, Baowen Zhang, Alex Trevino. An end-to-end Visium spatial transcriptomics computational pipeline for generating low-code interactive reports of spatial insights [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 2 (Clinical Trials and Late-Breaking Research); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(8_Suppl):Abstract nr LB079.

Batch Correction and Harmonization of -Omics Datasets with a Tunable Median Polish of Ratio.

Large scale -omics datasets can provide new insights into normal and disease-related biology when analyzed through a systems biology framework. However, technical artefacts present in most -omics datasets due to variations in sample preparation, batching, platform settings, personnel, and other experimental procedures prevent useful analyses of such data without prior adjustment for these technical factors. Here, we demonstrate a tunable median polish of ratio (TAMPOR) approach for batch effect correction and agglomeration of multiple, multi-batch, site-specific cohorts into a single analyte abundance data matrix that is suitable for systems biology analyses. We illustrate the utility and versatility of TAMPOR through four distinct use cases where the method has been applied to different proteomic datasets, some of which contain a specific defect that must be addressed prior to analysis. We compare quality control metrics and sources of variance before and after application of TAMPOR to show that TAMPOR is effective at removing batch effects and other unwanted sources of variance in -omics data. We also show how TAMPOR can be used to harmonize -omics datasets even when the data are acquired using different analytical approaches. TAMPOR is a powerful and flexible approach for cleaning and harmonization of -omics data prior to downstream systems biology analysis.

Abstract 2454: Identification of colorectal cancer stem cells from single cell RNA sequencing

Abstract Cancer stem cells (CSCs) play a critical role in metastasis, relapse and therapy resistance in colorectal cancer (CRC). Although the normal lineage of cell development in the intestine has identified many of the genes involved in the induction and maintenance of pluripotency, recent studies of CSCs suggest significant heterogeneity and lack of common surface markers for identification. In this study, we perform single cell RNA sequencing (scRNA-seq) to extract a single cell stemness signature (SCS_sig) for colorectal CSCs identification and as a potential predictor of relapse after surgical resection. A total of 26 fresh surgery specimens (6 colon tumor, 3 colon matched normal, 9 liver metastasis, 8 liver matched normal) from neoadjuvant chemotherapy (NAC) treated patients with single-cell transcriptome generated using 10X Chromium scRNA-seq and Illumina platforms were analyzed. Following extensive quality control and batch effect correction, we identified 20k epithelial cells using canonical epithelial cell marker genes (EPCAM, KRT8 and KRT18) from a total of 111,200 cells, with mean 1113 epithelial cell per sample (excluding liver normal samples).To obtain the CSCs profiling signature, we first annotated a ‘gold-standard’ set of CSC that expressed all canonical colon CSC marker genes (LGR5, ASCL2, EPHB2, PROM1, AXIN2, LEFTY1, RNF43, CD44 and SLC12A2), identifying 346 CSCs (1.71% of epithelial cells). Published stemness signatures were significantly enriched in the gold standard stem cells (false discovery rate (FDR) < 0.05). Then we selected the 50 most significantly up-regulated genes in this gold standard set (relative to other epithelial cells) as a single-cell stemness signature (named SCS_sig hereafter). Interesting, we saw that in every tumor there was a continuum of stemness, rather than distinct stem and differentiated populations. The stemness state by SCS_sig was positively correlated with less differentiated state calculated by CytoTRACE score (r = 0.76, p< 0.001), a computational method that predicts the differentiation state of cells from scRNA-seq data. The SCS_sig was significantly higher in tumor cells compared to normal cells in our cohort as well as the previously published SMC cohort (fold change = 2.52 and p < 0.001, fold change = 1.44 and p < 0.001, respectively). There was notable heterogeneity in the fraction of stem-like tumor cells with in 15 tumor samples, with the tumor with greatest CSC percentage having the shortest time to relapse. The SCS_sig generated from stem-like cells can identify the stemness state of epithelial cells in scRNAseq data. A signature score is more simple and robust than using few or panel of CSCs marker genes to identify single CSC because of the reads dropout in scRNA-seq. Further investigation of the SCS_sig as a predictive and prognostic biomarker in CRC is ongoing in larger patient cohorts. Citation Format: Kangyu Lin, Mohammad Zeineddine, Saikat Chowdhury, Nicholas Hornstein, Sendurai Mani, Scott Kopetz, John Paul Shen. Identification of colorectal cancer stem cells from single cell RNA sequencing [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 2454.

Structure-preserving visualization for single-cell RNA-Seq profiles using deep manifold transformation with batch-correction

Dimensionality reduction and visualization play an important role in biological data analysis, such as data interpretation of single-cell RNA sequences (scRNA-seq). It is desired to have a visualization method that can not only be applicable to various application scenarios, including cell clustering and trajectory inference, but also satisfy a variety of technical requirements, especially the ability to preserve inherent structure of data and handle with batch effects. However, no existing methods can accommodate these requirements in a unified framework. In this paper, we propose a general visualization method, deep visualization (DV), that possesses the ability to preserve inherent structure of data and handle batch effects and is applicable to a variety of datasets from different application domains and dataset scales. The method embeds a given dataset into a 2- or 3-dimensional visualization space, with either a Euclidean or hyperbolic metric depending on a specified task type with type static (at a time point) or dynamic (at a sequence of time points) scRNA-seq data, respectively. Specifically, DV learns a structure graph to describe the relationships between data samples, transforms the data into visualization space while preserving the geometric structure of the data and correcting batch effects in an end-to-end manner. The experimental results on nine datasets in complex tissue from human patients or animal development demonstrate the competitiveness of DV in discovering complex cellular relations, uncovering temporal trajectories, and addressing complex batch factors. We also provide a preliminary attempt to pre-train a DV model for visualization of new incoming data.

Abstract 2075: Highly customizable multi-sample single cell RNA-Seq pipeline on the CGC

Abstract Single-cell (sc) transcriptomics has revolutionized our understanding of the biological characteristics and dynamics of cancer development. It can help us identify rare cell subpopulations and understand mechanisms associated with tumor genesis, progression, and response to therapy. The most important step in the analyses of any scRNA-seq dataset is subpopulation identification, usually performed via unsupervised clustering, followed by gene marker identification. We created a highly customizable workflow for sc data analysis, implemented in Common Workflow Language (CWL) on the Cancer Genomics Cloud (CGC) platform. The NCI-funded CGC platform, powered by Seven Bridges, provides a collaborative cloud base computation infrastructure that collocates computation, over 750 bioinformatics workflows, and 3+ PB data to researchers, making the analysis of large datasets accessible from any environment. The “Multi-Sample Clustering and Gene Marker Identification with Seurat 4.1.0” workflow comprises the following steps: Loading scRNA-seq Expression Datasets, Quality Control and Preprocessing, and Clustering and Identification of Gene Markers. Our solution supports gene-cell count matrices generated by several commonly used quantifiers (for example, Cell Ranger counts, Salmon Alevin, Kallisto BUStools, STAR) from single or multiple sc datasets from different batches, as well as single or multiple single-cell samples combined in a single SingleCellExperiment object. The versatility of the pipeline is obtained using several implemented options in each of the steps. Quality control can be performed manually or automatically using several options for normalization (LogNormalize, Deconvolution, SCnorm and Linnorm) and for batch effect correction (Seurat and Harmony). For clustering, the pipeline uses Seurat's graph-based approach, with options for different clustering resolutions. After performing identification of gene markers for each cluster, a researcher can test differential expression using various packages including wilcox, bimod, roc, and DESeq2. Here, we demonstrate the application of this workflow to a typical sc analysis, by processing an open access dataset of 61k cells isolated from embryonal mouse pons and forebrain, two major brain tumor locations. We used different clustering resolutions to achieve different degrees of granularity and identified cluster-specific marker genes used to identify vulnerable cell populations. To enable researchers to use this analysis as a guideline, we made this analysis available as a public project. Further development of single-cell sequencing techniques will undoubtedly improve our understanding of tumor biology and highlight promising drug targets. CGC’s cloud base computation infrastructure, along with numerous available cancer datasets and easy-to-use single-cell data processing workflows, among others, will be instrumental in this process. Citation Format: Nevena Vukojicic, Aleksandar Danicic, Zelia Worman, Rowan Beck, Dalibor Veljkovic, Marko Matic, Jack DiGiovanna, Brandi Davis-Dusenbery. Highly customizable multi-sample single cell RNA-Seq pipeline on the CGC [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 2075.

Abstract 2093: Introducing InMoose, an integrated open source Python package for multi-omic analyses

Abstract The recent exponential progress of sequencing technologies has dramatically impacted cancer research and paved the way to precision medicine in cancer care. In parallel, light-speed progress in bioinformatics has been essential to allow analysts to embrace the vast amount of data yielded by high-throughput profiling machines, turn this data into cancer biology knowledge, and ultimately develop innovative approaches to cancer care. Still, computational complexity and tools' interoperability remain major challenges for the advancement of -omic data-driven cancer research. Despite the historical prevalence of R, Python is gaining momentum in the bioinformatics landscape. As a general purpose language, it offers numerous advantages: - its overall ecosystem facilitates the integration of bioinformatics tools into large-scale frameworks, increasing their versatility and widening their targeted audience; - its straightforward syntax and user-friendly logic makes it a popular language in education, and explains its wide adoption in various sectors; - its wide adoption favors interdisciplinarity, e.g. it can reduce the learning curve for engineers eager to make an impact in cancer research. These advantages motivate the further development of the Python bioinformatics ecosystem. We thus advocate the necessity to port reference tools from R to Python, with the ambition of shaping a comprehensive ecosystem. Sign of this trend, state-of-the-art tools have been directly developed in Python (e.g. lifelines, a library dedicated to survival analysis) or quickly ported in Python from R (e.g. harmony, an R library for integrating single cell data). We introduce InMoose (Integrated Multi-Omics Open Source Environment), an open source Python unified framework for every -omic data type. It is based on recognized tools and focuses on efficiency and user-friendliness. InMoose is accessible at https://github.com/epigenelabs/inmoose and is released under GPL3 license. The first version of InMoose focuses on bringing to the Python world transcriptomics tools mostly based on the edgeR R package. It features batch-effect correction algorithms, as well as differential expression analysis, for microarray and RNA-seq data. InMoose demonstrates the advantages of our approach: - it is developed with the intent to capitalize on and facilitate interactions with other popular Python data-oriented libraries (e.g. pandas, numpy); - it demonstrates significant computational performance improvements; - our package integrates easily into web-based user-friendly analyses platforms (e.g. Epigene Labs’ proprietary mCUBE platform); - porting existing code from one language to another provides an opportunity window to improve both functionality and performance. We expect that our effort will help foster a larger collaborative effort to build and grow a consistent state-of-the-art Python platform for cancer bioinformatics. Citation Format: Maximilien Colange, Guillaume Appé, Akpéli Nordor, Abdelkader Behdenna. Introducing InMoose, an integrated open source Python package for multi-omic analyses [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 2093.

Heuristic normalization procedure for batch effect correction

Heuristic normalization procedure for batch effect correction

Spatially informed clustering, integration, and deconvolution of spatial transcriptomics with GraphST

Spatial transcriptomics technologies generate gene expression profiles with spatial context, requiring spatially informed analysis tools for three key tasks, spatial clustering, multisample integration, and cell-type deconvolution. We present GraphST, a graph self-supervised contrastive learning method that fully exploits spatial transcriptomics data to outperform existing methods. It combines graph neural networks with self-supervised contrastive learning to learn informative and discriminative spot representations by minimizing the embedding distance between spatially adjacent spots and vice versa. We demonstrated GraphST on multiple tissue types and technology platforms. GraphST achieved 10% higher clustering accuracy and better delineated fine-grained tissue structures in brain and embryo tissues. GraphST is also the only method that can jointly analyze multiple tissue slices in vertical or horizontal integration while correcting batch effects. Lastly, GraphST demonstrated superior cell-type deconvolution to capture spatial niches like lymph node germinal centers and exhausted tumor infiltrating T cells in breast tumor tissue.

Rectal cancer–specific subtyping using transcriptomics datasets derived from treatment-naive patients to assess novel prognostic signatures and key differences to CMS-based subtyping.

235 Background: Molecular subtyping in colorectal cancer provided novel insights into tumour heterogeneity and delivered new prognostic signatures. Existing methods (such as CMS-based subtyping) were generated without much consideration of differences in expression profiles between colon and rectal cancer tissues. Moreover, stage 2 and 3 rectal cancer patients often receive neo-adjuvant radio-chemotherapy which may impact on gene expression when determined from resection samples. Methods: We investigated possible differences between treatment-naive colon and rectal cancers at the molecular level. We collected mRNA expression profiles for rectal and colon cancer samples available from Gene Expression Omnibus (GEO) database (n=2139) and applied batch-effect correction using ComBat. Results: To explore whether CMS subtypes changed after neoadjuvant chemoradiation in rectal cancers, we analysed rectal tumour samples from 40 patients where both pre-operative biopsies and resections after chemoradiotherapy were available. More than half of samples shifted from CMS1-3 to CMS4 after patients received chemoradiotherapy. We therefore focused our further analysis on treatment-naive rectal cancers. We observed that CMS-based subtyping showed different prognostic potential in treatment-naïve colon vs. rectal cancers. Rectal cancers had significantly better disease-free survival (DFS) when assigned to CMS 2 or 4 compared to colon cancers. This suggested differences in disease biology that were not captured by CMS subtyping. We next aimed to identify molecular subtypes in rectal cancers, and identified three prevalent rectal-specific subtypes (RSS) by applying hierarchical clustering in 182 untreated rectal cancer samples. We used gene co-expression network analysis to define functional gene modules to identify these subtypes. Based on the gene modules’ signatures, we built a classifier to predict our new subtypes in different transcriptomics datasets (n=692). We observed that RSS1 had better disease-free survival (DFS), whereas RSS2 exhibited worse short-term DFS (less than 3 years) and RSS3 showed worse long-term DFS (3+ years). We discovered the deactivation of EGFR, MAPK and WNT pathways on RSS3 as well as high CD8+ infiltration and low CD4+ counts in these tumours. RSS2 exhibits low plasma cell and regulatory T-cell abundance and high activations of TGF-β, NF-κB, and TNF-α. We also observed high EMT, angiogenesis and inflammatory response in this subtype. Lastly, RSS1 showed high expression on MYC target and low activity on angiogenesis genes. Moreover, the immune and inflammatory responses were also lower than in the other subtypes. Conclusions: Rectal-specific subtypes are of prognostic importance and provide new insight into molecular disease pathways and potential targets.