2 October 2021 in Bioinformatics
Genome‐wide association studies (GWAS) summary statistics have popularized and accelerated genetic research. However, a lack of standardization of the file formats used has proven problematic when running secondary analysis tools or performing meta-analysis studies. To address this issue, we have developed MungeSumstats, a Bioconductor R package for the standardization and quality control of GWAS summary statistics. MungeSumstats can handle the most common summary statistic formats, including variant call format (VCF) producing a reformatted, standardized, tabular summary statistic file, VCF or R native data object. MungeSumstats is available on Bioconductor (v 3.13) and can also be found on Github at: https://neurogenomics.github.io/MungeSumstats. Supplementary data are available at Bioinformatics online.
22 December 2022 in Nature Communications
Recently, Zimmerman et al. 1 , highlighted the importance of accounting for the dependence between cells from the same individual when conducting differential expression analysis on single-cell RNAsequencing data. Their work proved the inadequacy of pseudoreplication approaches for such analysis-this was an important step forward that was conclusively proven by them. However, there appear to be limitations in both their benchmarking and simulation approaches. Here, we corrected these issues, reran the author's analysis and found that pseudobulk methods outperformed mixed models. Based on these findings, we recommend the use of pseudobulk approaches for differential expression in single-cell RNA-sequencing analyses. Zimmerman et al. 1 , performed a systematic analysis of differential expression methods' type-1 error rates; the proportion of nondifferentially expressed genes indicated as differentially expressed by a model. Their analysis was conducted on simulated, single-cell expression data across 20,000 iterations. The authors tested iterations of 5-40 individuals and 50-500 cells using an unadjusted p-value cutoff of 0.05 for significance. Plotting the results showed pseudobulk approaches had the lowest type-1 error at every iteration (Supplementary Fig. 1). However, evaluating such models on their type-1 or type-2 error rate in isolation is insufficient to determine their true performance. For example, a method with low type-1 error may have a high type-2 error rate. Therefore, we need to consider both type-1 and type-2 error rate to accurately benchmark the models. Moreover, because no seed was set for the pseudo-random number generator used in their hierarchical single-cell expression simulation approach (hierarchicell), the different methods evaluated by Zimmerman et al. were compared on different simulated datasets. Here, we modified Zimmerman et al.'s hierachicell approach to simulate both differentially expressed and non-differentially expressed genes. The differentially expressed genes were randomly simulated with a fold change between 1.1 and 10. We further modified hierachicell to correct the seeding of the pseudo-random number generator to enable fair comparisons across models. We tested the models using the Matthews Correlation Coefficient (MCC) giving a balanced measure of performance. MCC is a
19 August 2021 in bioRxiv
Advances in single-cell RNA-sequencing technology over the last decade have enabled exponential increases in throughput: datasets with over a million cells are becoming commonplace. The burgeoning scale of data generation, combined with the proliferation of alternative analysis methods, led us to develop the scFlow toolkit and the nf-core/scflow pipeline for reproducible, efficient, and scalable analyses of single-cell and single-nuclei RNA-sequencing data. The scFlow toolkit provides a higher level of abstraction on top of popular single-cell packages within an R ecosystem, while the nf-core/scflow Nextflow pipeline is built within the nf-core framework to enable compute infrastructure-independent deployment across all institutions and research facilities. Here we present our flexible pipeline, which leverages the advantages of containerization and the potential of Cloud computing for easy orchestration and scaling of the analysis of large case/control datasets by even non-expert users. We demonstrate the functionality of the analysis pipeline from sparse-matrix quality control through to insight discovery with examples of analysis of four recently published public datasets and describe the extensibility of scFlow as a modular, open-source tool for single-cell and single nuclei bioinformatic analyses.
16 November 2024 in Nature Communications
Understanding how genetic variants affect the epigenome is key to interpreting GWAS, yet profiling these effects across the non-coding genome remains challenging due to experimental scalability. This necessitates accurate computational models. Existing machine learning approaches, while progressively improving, are confined to the cell types they were trained on, limiting their applicability. Here, we introduce Enformer Celltyping, a deep learning model which incorporates distal effects of DNA interactions, up to 100,000 base-pairs away, to predict epigenetic signals in previously unseen cell types. Using DNA and chromatin accessibility data for epigenetic imputation, Enformer Celltyping outperforms current best-in-class approaches and generalises across cell types and biological regions. Moreover, we propose a framework for evaluating models on genetic variant effect prediction using regulatory quantitative trait loci mapping studies, highlighting current limitations in genomic deep learning models. Despite this, Enformer Celltyping can also be used to study cell type-specific genetic enrichment of complex traits.
11 December 2024 in Nucleic Acids Research
To understand the complex relationship between histone mark activity and gene expression, recent advances have used in silico predictions based on large-scale machine learning models. However, these approaches have omitted key contributing factors like cell state, histone mark function or distal effects, which impact the relationship, limiting their findings. Moreover, downstream use of these models for new biological insight is lacking. Here, we present the most comprehensive study of this relationship to date – investigating seven histone marks in eleven cell types across a diverse range of cell states. We used convolutional and attention-based models to predict transcription from histone mark activity at promoters and distal regulatory elements. Our work shows that histone mark function, genomic distance and cellular states collectively influence a histone mark’s relationship with transcription. We found that no individual histone mark is consistently the strongest predictor of gene expression across all genomic and cellular contexts. This highlights the need to consider all three factors when determining the effect of histone mark activity on transcriptional state. Furthermore, we conducted in silico histone mark perturbation assays, uncovering functional and disease related loci and highlighting frameworks for the use of chromatin deep learning models to uncover new biological insight.
31 December 2021 in Molecular & Cellular Proteomics
The amount of any given protein in the brain is determined by the rates of its synthesis and destruction, which are regulated by different cellular mechanisms. Here, we combine metabolic labeling in live mice with global proteomic profiling to simultaneously quantify both the flux and amount of proteins in mouse models of neurodegeneration. In multiple models, protein turnover increases were associated with increasing pathology. This method distinguishes changes in protein expression mediated by synthesis from those mediated by degradation. In the AppNL-F knockin mouse model of Alzheimer’s disease, increased turnover resulted from imbalances in both synthesis and degradation, converging on proteins associated with synaptic vesicle recycling (Dnm1, Cltc, Rims1) and mitochondria (Fis1, Ndufv1). In contrast to disease models, aging in wild-type mice caused a widespread decrease in protein recycling associated with a decrease in autophagic flux. Overall, this simple multidimensional approach enables a comprehensive mapping of proteome dynamics and identifies affected proteins in mouse models of disease and other live animal test settings.