Blind estimation and correction of microarray batch effect.

In addition to oncogenic 'high-risk' human papillomaviruses (HR-HPV), several co-factors are needed in cervical carcinogenesis, but it is poorly understood whether these HPV co-factors associated with incident cervical intraepithelial neoplasia (CIN) grade 1 are different from those required for progression to CIN2 and CIN3. To gain further insights into the true biological differences between CIN1, CIN2 and CIN3, we assessed HPV co-factors increasing the risk of incident CIN1, CIN2 and CIN3. Data from the New Independent States of the Former Soviet Union (NIS) Cohort (n = 3187) and the Latin American Screening (LAMS) Study (n = 12,114) were combined, and co-factors associated with progression to CIN1, CIN2 and CIN3 were analysed using multinomial logistic regression models with all covariates recorded at baseline. HR-HPV-positive women (n = 1105) represented a subcohort of all 1865 women prospectively followed up in both studies. Altogether, 90 (4.8%), 39 (2.1%) and 14 (1.4%) cases progressed to CIN1, CIN2 and CIN3, respectively. Baseline HR-HPV was the single most powerful predictor of incident CIN1, CIN2 and CIN3. When controlled for residual HPV confounding by analysing HR-HPV-positive women only, the risk profiles of incident CIN1, CIN2 and CIN3 were unique. Completely different HPV co-factors were associated with progression to CIN1, CIN2 and CIN3 in univariate and multivariate analyses, irrespective of whether non-progression, CIN1 or CIN2 was used as the reference outcome. HPV co-factors associated with progression to CIN1, CIN2 and CIN3 display unique profiles, implicating genuine biological differences between the three CIN grades, which prompts us to re-visit the concept of combining CIN2 with CIN3 or CIN1.

MotivationBatch effects in omics datasets are usually a source of technical noise that masks the biological signal and hampers data analysis. Batch effect removal has been widely addressed for individual omics technologies. However, multi-omic datasets may combine data obtained in different batches where omics type and batch are often confounded. Moreover, systematic biases may be introduced without notice during data acquisition, which creates a hidden batch effect. Current methods fail to address batch effect correction in these cases.ResultsIn this article, we introduce the MultiBaC R package, a tool for batch effect removal in multi-omics and hidden batch effect scenarios. The package includes a diversity of graphical outputs for model validation and assessment of the batch effect correction.Availability and implementationMultiBaC package is available on Bioconductor (https://www.bioconductor.org/packages/release/bioc/html/MultiBaC.html) and GitHub (https://github.com/ConesaLab/MultiBaC.git). The data underlying this article are available in Gene Expression Omnibus repository (accession numbers GSE11521, GSE1002, GSE56622 and GSE43747).Supplementary informationSupplementary data are available at Bioinformatics online.

BackgroundHigh throughput RNA sequencing is a powerful approach to study gene expression. Due to the complex multiple-steps protocols in data acquisition, extreme deviation of a sample from samples of the same treatment group may occur due to technical variation or true biological differences. The high-dimensionality of the data with few biological replicates make it challenging to accurately detect those samples, and this issue is not well studied in the literature currently. Robust statistics is a family of theories and techniques aim to detect the outliers by first fitting the majority of the data and then flagging data points that deviate from it. Robust statistics have been widely used in multivariate data analysis for outlier detection in chemometrics and engineering. Here we apply robust statistics on RNA-seq data analysis.ResultsWe report the use of two robust principal component analysis (rPCA) methods, PcaHubert and PcaGrid, to detect outlier samples in multiple simulated and real biological RNA-seq data sets with positive control outlier samples. PcaGrid achieved 100% sensitivity and 100% specificity in all the tests using positive control outliers with varying degrees of divergence. We applied rPCA methods and classical principal component analysis (cPCA) on an RNA-Seq data set profiling gene expression of the external granule layer in the cerebellum of control and conditional SnoN knockout mice. Both rPCA methods detected the same two outlier samples but cPCA failed to detect any. We performed differentially expressed gene detection before and after outlier removal as well as with and without batch effect modeling. We validated gene expression changes using quantitative reverse transcription PCR and used the result as reference to compare the performance of eight different data analysis strategies. Removing outliers without batch effect modeling performed the best in term of detecting biologically relevant differentially expressed genes.ConclusionsrPCA implemented in the PcaGrid function is an accurate and objective method to detect outlier samples. It is well suited for high-dimensional data with small sample sizes like RNA-seq data. Outlier removal can significantly improve the performance of differential gene detection and downstream functional analysis.

Cross-species single-cell RNA-seq data hold immense potential for unraveling cell type evolution and transferring knowledge between well-explored and less-studied species. However, challenges arise from interspecific genetic variation, batch effects stemming from experimental discrepancies and inherent individual biological differences. Here, we benchmarked nine data-integration methods across 20 species, encompassing 4.7 million cells, spanning eight phyla and the entire animal taxonomic hierarchy. Our evaluation reveals notable differences between the methods in removing batch effects and preserving biological variance across taxonomic distances. Methods that effectively leverage gene sequence information capture underlying biological variances, while generative model-based approaches excel in batch effect removal. SATURN demonstrates robust performance across diverse taxonomic levels, from cross-genus to cross-phylum, emphasizing its versatility. SAMap excels in integrating species beyond the cross-family level, especially for atlas-level cross-species integration, while scGen shines within or below the cross-class hierarchy. As a result, our analysis offers recommendations and guidelines for selecting suitable integration methods, enhancing cross-species single-cell RNA-seq analyses and advancing algorithm development.

Background: Community acquired pneumonia (CAP) is a major global health concern for children, causing numerous hospitalizations and death. Community acquired pneumonia is a major cause of under-five mortality in children. Identifying reliable prognostic markers is crucial. Mean platelet volume (MPV) and red cell distribution width co-efficient of variation (RDW-CV) are accessible and cost-effective options for assessment of prognosis in pneumonia. This study investigates MPV and RDW-CV as prognostic markers in children with community acquired pneumonia. Methods: This prospective observational study includes 80 children aged 1-16 years diagnosed with pneumonia upon initial examination and admitted to hospital. Community acquired Pneumonia diagnosis was based on clinical symptoms, physical examination and/or radiological findings. Pneumonia severity was assessed using the Clinical Respiratory Score (CRS), categorizing patients into mild, moderate, severe groups. MPV and RDW-CV were compared among these groups. Results: Across the sample of 80 children, RDW-CV and MPV values did not demonstrate statistically significant differences when compared by sex, age group, ICU requirement or hospital stay duration. For most comparisons, p-values exceeded 0.05, indicating that variations in these indices are likely due to random fluctuations rather than true biological differences. Conclusions: Certain trends (higher RDW-CV and MPV in intense severity groups, elevated MPV in pleural effusion and significant ROC for MPV) are biologically plausible in the context of systemic inflammation. MPV appears to be a more reliable marker than RDW-CV in this dataset. RDW-CV did not demonstrate meaningful prognostic value. The findings reinforce that MPV may serve as a supportive, accessible biomarker for predicting more severe pneumonia in children, whereas RDW-CV requires larger sample evaluation to confirm its utility.

BackgroundThe brain‐age algorithm utilizes a regression model of the typical aging trajectory, trained on structural MRIs and chronological ages of a large population of healthy individuals spanning a wide age range (Cole et al. 2017, Neuroimage). MRIs from new populations are then compared to this regression for estimating brain age, a biomarker of brain health. Few studies have investigated stability of the brain‐age model across field strength and the influence of voxel size and race is understudied.MethodWe analyzed stability of the brain‐age model within and across field strength in the ADNI cohort (n = 75: 25 CN/25 MCI/25 AD) using intra‐class correlation coefficients (ICC) and multiple R2 values. Impact of voxel size (0.8mm, 1mm, and ADNI2 sequence 1.1×1.1×1.2mm) was compared in a sample from the Indiana ADRC (n = 48: 19 CN/16 SCD/10 MCI/3 AD). We then calculated brain age gap (BAG; estimated– chronological age) and compared results across self‐reported racial and ethnic groups matched on diagnosis, sex, age, education, and handedness using independent t‐tests. BAG results of African Americans (AAs) from the IADRC (n = 71: 30 CN/19 SCD/15 MCI/7 AD) and ADNI (n = 147: 41 CN/20 EMCI/27 LMCI/15 AD/44 SMC), and Hispanics (n = 76: 24 CN/16 EMCI/11 LMCI/11 AD/14 SMC) and Asians (n = 47: 12 CN/7 EMCI/4 LMCI/9 AD/15 SMC) from ADNI only, were compared to matched non‐Hispanic Whites (NHWs).ResultThe brain‐age method was stable across repeated scans within field strength, but less stable across field strengths (Fig.1). 3T scans consistently estimated individuals to be ∼4.43 years younger than the 1.5T scans. The method was relatively stable within and across voxel sizes (Fig.2). There was a significant difference in BAG between races (NHWs and AAs, and NHWs and Asians), but not between ethnicities (NHWs and Hispanics) (Fig.3).ConclusionDue to the model’s variability across field strength, we propose a linear offset equation for data harmonization (Fig.1): ((3T estimate/0.92)+9.55). We are also investigating ComBat harmonization to eliminate effects of field strength. While other studies show structural differences across race, it will be important to determine whether the self‐reported racial differences in BAG are due to inherent model bias or true biological differences.

BackgroundThe brain‐age algorithm utilizes a regression model of the typical aging trajectory, trained on structural MRIs and chronological ages of a large population of healthy individuals spanning a wide age range (Cole et al. 2017, Neuroimage). MRIs from new populations are then compared to this regression for estimating brain age, a biomarker of brain health. Few studies have investigated stability of the brain‐age model across field strength and the influence of voxel size and race is understudied.MethodWe analyzed stability of the brain‐age model within and across field strength in the ADNI cohort (n = 75: 25 CN/25 MCI/25 AD) using intra‐class correlation coefficients (ICC) and multiple R2 values. Impact of voxel size (0.8mm, 1mm, and ADNI2 sequence 1.1×1.1×1.2mm) was compared in a sample from the Indiana ADRC (n = 48: 19 CN/16 SCD/10 MCI/3 AD). We then calculated brain age gap (BAG; estimated– chronological age) and compared results across self‐reported racial and ethnic groups matched on diagnosis, sex, age, education, and handedness using independent t‐tests. BAG results of African Americans (AAs) from the IADRC (n = 71: 30 CN/19 SCD/15 MCI/7 AD) and ADNI (n = 147: 41 CN/20 EMCI/27 LMCI/15 AD/44 SMC), and Hispanics (n = 76: 24 CN/16 EMCI/11 LMCI/11 AD/14 SMC) and Asians (n = 47: 12 CN/7 EMCI/4 LMCI/9 AD/15 SMC) from ADNI only, were compared to matched non‐Hispanic Whites (NHWs).ResultThe brain‐age method was stable across repeated scans within field strength, but less stable across field strengths (Fig.1). 3T scans consistently estimated individuals to be ∼4.43 years younger than the 1.5T scans. The method was relatively stable within and across voxel sizes (Fig.2). There was a significant difference in BAG between races (NHWs and AAs, and NHWs and Asians), but not between ethnicities (NHWs and Hispanics) (Fig.3).ConclusionDue to the model’s variability across field strength, we propose a linear offset equation for data harmonization (Fig.1): ((3T estimate/0.92)+9.55). We are also investigating ComBat harmonization to eliminate effects of field strength. While other studies show structural differences across race, it will be important to determine whether the self‐reported racial differences in BAG are due to inherent model bias or true biological differences.

BackgroundIn proteomics, batch effects are technical sources of variation that confounds proper analysis, preventing effective deployment in clinical and translational research.ResultsUsing simulated and real data, we demonstrate existing batch effect-correction methods do not always eradicate all batch effects. Worse still, they may alter data integrity, and introduce false positives. Moreover, although Principal component analysis (PCA) is commonly used for detecting batch effects. The principal components (PCs) themselves may be used as differential features, from which relevant differential proteins may be effectively traced. Batch effect are removable by identifying PCs highly correlated with batch but not class effect.However, neither PC-based nor existing batch effect-correction methods address well subtle batch effects, which are difficult to eradicate, and involve data transformation and/or projection which is error-prone. To address this, we introduce the concept of batch-effect resistant methods and demonstrate how such methods incorporating protein complexes are particularly resistant to batch effect without compromising data integrity.ConclusionsProtein complex-based analyses are powerful, offering unparalleled differential protein-selection reproducibility and high prediction accuracy. We demonstrate for the first time their innate resistance against batch effects, even subtle ones. As complex-based analyses require no prior data transformation (e.g. batch-effect correction), data integrity is protected. Individual checks on top-ranked protein complexes confirm strong association with phenotype classes and not batch. Therefore, the constituent proteins of these complexes are more likely to be clinically relevant.

Decreasing costs are making it feasible to perform time series proteomics and genomics experiments with more replicates and higher resolution than ever before. With more replicates and time points, proteome and genome-wide patterns of expression are more readily discernible. These larger experiments require more batches exacerbating batch effects and increasing the number of bias trends. In the case of proteomics, where methods frequently result in missing data this increasing scale is also decreasing the number of peptides observed in all samples. The sources of batch effects and missing data are incompletely understood necessitating novel techniques. Here we show that by exploiting the structure of time series experiments, it is possible to accurately and reproducibly model and remove batch effects. We implement Learning and Imputation for Mass-spec Bias Reduction (LIMBR) software, which builds on previous block-based models of batch effects and includes features specific to time series and circadian studies. To aid in the analysis of time series proteomics experiments, which are often plagued with missing data points, we also integrate an imputation system. By building LIMBR for imputation and time series tailored bias modeling into one straightforward software package, we expect that the quality and ease of large-scale proteomics and genomics time series experiments will be significantly increased. Python code and documentation is available for download at https://github.com/aleccrowell/LIMBR and LIMBR can be downloaded and installed with dependencies using 'pip install limbr'. Supplementary data are available at Bioinformatics online.

Based on benchmarking various methods and analyzing multiple real scRNA-seq datasets, a computational platform/workflow and a set of tips for best practices are developed for analyzing causal interactions or relationships in single cells.

BackgroundCell composition deconvolution (CCD) is a type of bioinformatic task to estimate the cell fractions from bulk gene expression profiles, such as RNA-seq. Many CCD models were developed to perform linear regression analysis using reference gene expression signatures of distinct cell types. Reference gene expression signatures could be generated from cell-specific gene expression profiles, such as scRNA-seq. However, the batch effects and dropout events frequently observed across scRNA-seq datasets have limited the performances of CCD methods.MethodsWe developed a deep neural network (DNN) model, HASCAD, to predict the cell fractions of up to 15 immune cell types. HASCAD was trained using the bulk RNA-seq simulated from three scRNA-seq datasets that have been normalized by using a Harmony-Symphony based strategy. Mean square error and Pearson correlation coefficient were used to compare the performance of HASCAD with those of other widely used CCD methods. Two types of datasets, including a set of simulated bulk RNA-seq, and three human PBMC RNA-seq datasets, were arranged to conduct the benchmarks.ResultsHASCAD is useful for the investigation of the impacts of immune cell heterogeneity on the therapeutic effects of immune checkpoint inhibitors, since the target cell types include the ones known to play a role in anti-tumor immunity, such as three subtypes of CD8 T cells and three subtypes of CD4 T cells. We found that the removal of batch effects in the reference scRNA-seq datasets could benefit the task of CCD. Our benchmarks showed that HASCAD is more suitable for analyzing bulk RNA-seq data, compared with the two widely used CCD methods, CIBERSORTx and quanTIseq. We applied HASCAD to analyze the liver cancer samples of TCGA-LIHC, and found that there were significant associations of the predicted abundance of Treg and effector CD8 T cell with patients’ overall survival.ConclusionHASCAD could predict the cell composition of the PBMC bulk RNA-seq and classify the cell type from pure bulk RNA-seq. The model of HASCAD is available at https://github.com/holiday01/HASCAD.

Removal of batch effects using stratified subsampling of metabolomic data for in vitro endocrine disruptors screening

Genomic data integration is a key goal to be achieved towards large-scale genomic data analysis. This process is very challenging due to the diverse sources of information resulting from genomics experiments. In this work, we review methods designed to combine genomic data recorded from microarray gene expression (MAGE) experiments. It has been acknowledged that the main source of variation between different MAGE datasets is due to the so-called 'batch effects'. The methods reviewed here perform data integration by removing (or more precisely attempting to remove) the unwanted variation associated with batch effects. They are presented in a unified framework together with a wide range of evaluation tools, which are mandatory in assessing the efficiency and the quality of the data integration process. We provide a systematic description of the MAGE data integration methodology together with some basic recommendation to help the users in choosing the appropriate tools to integrate MAGE data for large-scale analysis; and also how to evaluate them from different perspectives in order to quantify their efficiency. All genomic data used in this study for illustration purposes were retrieved from InSilicoDB http://insilico.ulb.ac.be.

Bulk transcriptomes are an essential data resource for understanding basic and disease biology. However, integrating information from different experiments remains challenging because of the batch effect generated by various technological and biological variations in the transcriptome. Numerous batch-correction methods to deal with this batch effect have been developed in the past. However, a user-friendly workflow to select the most appropriate batch-correction method for the given set of experiments is still missing. We present the SelectBCM tool that prioritizes the most appropriate batch-correction method for a given set of bulk transcriptomic experiments, improving biological clustering and gene differential expression analysis. We demonstrate the applicability of the SelectBCM tool on analyses of real data for two common diseases, rheumatoid arthritis and osteoarthritis, and one example to characterize a biological state, where we performed a meta-analysis of the macrophage activation state. The R package is available at https://github.com/ebi-gene-expression-group/selectBCM.

Clustering and cell type classification are a vital step of analyzing scRNA-seq data to reveal the complexity of the tissue (e.g. the number of cell types and the transcription characteristics of the respective cell type). Recently, deep learning-based single-cell clustering algorithms become popular since they integrate the dimensionality reduction with clustering. But these methods still have unstable clustering effects for the scRNA-seq datasets with high dropouts or noise. In this study, a novel single-cell RNA-seq deep embedding clustering via convolutional autoencoder embedding and soft K-means (scCAEs) is proposed by simultaneously learning the feature representation and clustering. It integrates the deep learning with convolutional autoencoder to characterize scRNA-seq data and proposes a regularized soft K-means algorithm to cluster cell populations in a learned latent space. Next, a novel constraint is introduced to the clustering objective function to iteratively optimize the clustering results, and more importantly, it is theoretically proved that this objective function optimization ensures the convergence. Moreover, it adds the reconstruction loss to the objective function combining the dimensionality reduction with clustering to find a more suitable embedding space for clustering. The proposed method is validated on a variety of datasets, in which the number of clusters in the mentioned datasets ranges from 4 to 46, and the number of cells ranges from 90 to 30302. The experimental results show that scCAEs is superior to other state-of-the-art methods on the mentioned datasets, and it also keeps the satisfying compatibility and robustness. In addition, for single-cell datasets with the batch effects, scCAEs can ensure the cell separation while removing batch effects.