• enables global snapshot of transcriptional profiles with single-cell resolution

Brief history of transcriptional profiling:

• 2000s, microarrays enabled systematic measurements of transcriptional changes → mRNAs labeled, loaded onto chip where each spot on chip was coated to catch mRNA of specific gene → color of spot indicates ratio of transcript
• Bulk RNA sequencing made it much more feasible which transcriptional differences could be measured
  • quantification of gene expression
  • comparitive transcriptomics
• limitations involve the inability to resolve heterogeneity
• single cell also gives you dynamics of gene expression and cell identity, primarily because it can help you look at cell topology
• goals of scRNA-seq are 1) measure distribution of expression levels for each gene across population of cells 2) measure transcirptional differnces across and within groups of cells 3) resolve single-cell heterogenity
• applications involve characterization of cell and transcriptional composition of tissues (cell type changes and their correlatoin)
• evaluate developmental processes (see expression changes while undergoing differentiation)
• evaluate how cancer evolution leads to mechanisms of therapeutic resistance

Droplet-based approach:

1. Tissue disassociated and samples created to separate individual cells, need to make sure tissue maintains integrity to prevent digestion of extracellular layers
2. samples collected, separated into individual droplets (use cytometry to evaluate cell viability through FACS to measure cell composition by pre-binding cell with fluorescent or antibodies to find expectation of proportion of cell types)
3. oil and water reaction has a single bead with barcode that bind to the specific mRNA of a single cell + done through cell where water and oils are displaced → you have a poisson distribution of beads and cells (proportion of concentration is based on concentration of cells and beads)
4. cells are lysed, reverse transcriptase synthesizes cDNA and applied molecular identity, use pCR and then add sequencing adapters for sequencing and assembly (3'-5')
   • lysis part is flawed right now because mRNA molecules don't always bind to the beads, RT enzymes are known to be inefficient, and RT + PCR depend on specific transcript, making this more variable

• onyl about 5% of mRNA is captured → UMI: total count of unique molculare identifies which are labelled on each mRNA with a unique barcode at the RT step
  • UMI reduces amplification error → evaluation of performance found that 10x is one of the better ones → SUPeR-seq wth Accuracy of 0.95, 4 million copies of mRNA sequence in every cell

Limitations:

• low capture probability and data sparsity is a big problem → you can use BDRhapsody to limit the scope of experiment to a few hundred mRNAs

• inability to analyze full-length transcriptomes → can use smartSEQ2

• inability to resovle spatial information → where computation comes into play

• integrating with measurements in Microscopy, FACs, total_Seq, hashtag tech, etc.

• preparing scRNAseq data for clustering → preprocessing: align and count UMIS

• finding feature selection and do dimensionality reduction → you can visualize hereogeneity with t-SNE and UMAP

• HVGs for preprocessing:

  • distinguish technical noise by looking at distribution of population of cells → HVGs are identified because they are interesting biological markers
  • PCA is then applied as a result
    • visualized using UMAP and t-SNE

• Quality control → normalization → feauture selection → dimensionality reduction → cell-cell distances → unsupervised clustering