Quantitative Biomedicine ✨

This repository houses conceptual viewpoints, coding practice, assignment/competition solutions based on the materials from a variety of computational biology/bioinformatics courses, workshops, technical manuals, academic articles, and others.

Last updated: 7 Feb 2025

Features

• Bulk RNA-seq data analysis
• Single cell RNA-seq data analysis
• ATAC-seq data analysis
• Multi-omics idea
• COVID-19 RNA-seq data resources

Technical procedures

Analyze bulk RNA-seq data

• Run FastQC or fastp to evaluate sequence quality and content
• [Recommend] Use splice-aware genome aligner STAR to align the reads
  • Other splice-aware alignment tool options: Olego, HISAT2, MapSplice, ABMapper, Passion, BLAT, RUM ...
  • Other alignment tools that disregard isoforms: BWA, Bowtie2 ...
• Use Rsubread to align the reads
  • Why align? To pinpoint the specific location on the human genome from which our reads originated
• Use Qualimap to perform quality assurance on the aligned reads
• Use MultiQC to harmonize all QC and alignment metadata from FastQC, STAR, Qualimap, and other tools
• Use GenomicAlignments for aligned reads to obtain the gene-level or exon-level quantification
• Use featureCounts for aligned reads to count the fragments
• [Recommend] Use Salmon for unaligned reads to obtain the transcript-level quantification
  • Why unalign? To speed up the counting process of reads
  • Next step: Use tximport to aggregate transcript-level quantification to the gene level
• Perform differential gene expression analysis
• Perform principal component analysis, heatmap, and clustering
• Perform gene set enrichment analysis
• Achieve cell-type resolution in bulk RNA-Seq through deconvolution techniques (Under Active Construction)

---

Analyze single cell RNA-seq data

• A mini scRNA-seq pipeline | Why apply pipelines?
• If given raw bcl files, we convert them to fastq files
• As inputs are fastq files, we can ...
  • Run FastQC to evaluate sequence quality and content
  • Use Trim Galore to trim reads if we spot unexpected low-quality base calls/adaptor contamination
  • Re-run FastQC to re-evaluate sequence quality and content
  • If single-cell RNA-seq data is generated from the plate-based protocol, we can ...
    • Use STAR to perform alignment and FeatureCounts to generate the count matrix
  • Else if single-cell RNA-seq data is generated from the droplet-based protocol, we can ...
    • Use kb-python package to perform pseudo sequence alignment and generate the count matrix
    • Use Cell Ranger pipelines to perform sequence alignment and generate the count matrix
• After having the feature-barcode matrices at hand, we can ...
  • Use Scanpy workflow to perform quality assurance, cell clustering, marker gene detection for cell identities, and trajectory inference
  • Use Seurat workflow to perform quality assurance, cell clustering, and marker gene detection for cell identities case 1 | case 2
    • If we observe the factor-specific clustering and want cells of the same cell type cluster together across single/multiple confounding factors, we can use canonical correlation analysis or Harmony (suitable for complicated confounding effects) to integrate cells
    • We can leverage SingleR or ScType to partially or fully automate cell-type identification
      • Other options of automating cell-type identification by mapping to references and then transfering labels: scArches, Symphony
  • Use Bioconductor packages to perform single cell RNA-Seq data analysis
  • Generate pseudobulk, which aggregates the gene expression levels specific to each cell type within an individual
  • Perform pseudobulk-based differentially gene expression analysis in edgeR or DESeq2
  • Use bulk RNAseq-based pathway analysis tools (e.g., clusterProfiler, GSEA, GSVA) or single cell RNAseq-based Pagoda2 to evaluate if a predefined set of genes shows statistically significant and consistent variations between biological conditions
  • Use scGen to model perturbation responses
• Frontier single-cell RNA-seq data analyses

---

Analyze ATAC-seq data

• Practical guide
• Run ENCODE ATAC-seq pipeline to perform alignment, quality assurance, peaking calling, and signal track generation
• If we're interested in inspecting every step in each analytical phase, or even leveraging advanced/unique features of other tools that the current pipeline ignores, ...
  • For alignment and post-alignment phases, we can ...
    • Use Rsubread or Rbowtie2 to align the fastq files relative to hg19/hg38/hs1
    • Use GenomicAlignments and GenomicRanges to perform post-alignment processing including reading properly paired reads, estimating MapQ scores/insert sizes, reconstructing the full-length fragment, and others
    • Use ATACseqQC to perform comprehensive ATAC-seq quality assurance
  • For TSS analysis phase, we can ...
    • Use soGGi to assess the transcriptional start site signal in the nucleosome-free open region
  • For peaking calling phase, we can ...
    • Use MACS2 and ChIPQC to call peaks in the nucleosome-free open region, and perform quality assurance
    • Or use Genrich to call peaks in the nucleosome-free open region
    • Or use MACS3/MACSr (R wrapper of MACS3) to call peaks in the nucleosome-free open region
    • Use ChIPseeker to annotate peak regions with genomic features
  • For functional analysis phase, we can ...
    • Use rGREAT to functionally interpret the peak regions based on the GO database
    • Use GenomicRanges and GenomicAlignments to select and count non-redundant peaks
    • Use DESeq2/DESeq2-based DiffBind and ChIPseeker to analyze differences in peaks with gene annotations across conditions
    • Use clusterProfiler to perform enrichment analysis of differential peak regions
    • However, functional insights gained by peak annotations can hardly illustrate what key regulators shape the transcription mechanism.
  • To further infer transcription factors acting in peak regions, we can ...
    • Use MotifDb/JASPAR2022 and seqLogo/ [recommend] ggseqlogo to search and visualize motifs
    • Use motifmatchr (R wrapper of MOODS) to map peaks to motifs, DNA sequences preferred by transcription factors
    • Use chromVAR to analyze differences in motifs across conditions
• Transfer the cell type labels from single-cell RNA-seq data to separately collected single-cell ATAC-seq data

---

Analyze proteomics data

• A quick start from loading an online spectrum, performing peak quality control, annotating peaks, to visualizing the annotated peaks

---

Conceptual lens

• High level multi-omics idea
• Case-control design
• Graph neural networks in biomedical studies