adv-single-cell-01.Rmd

---
title: "Large-scale clustering in Bioconductor"
description: >
  Learn how to do unsupervised clustering in R/Bioconductor for large-scale data.
author: "Stephanie Hicks"
output:
  rmarkdown::html_document:
   highlight: pygments
   toc: true
   toc_depth: 3
   fig_width: 5
vignette: >
  %\VignetteIndexEntry{Large-scale clustering in Bioconductor}
  %\VignetteEncoding[ut8]{inputenc}
  %\VignetteEngine{knitr::rmarkdown}
---


```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE, message = FALSE, warning = FALSE)
```


# Overview

## Key resources

- Workshop material: [pkgdown website](https://stephaniehicks.com/cshlgsd2022)
- Code: [GitHub](https://github.com/stephaniehicks/cshlgsd2022)


# Part 1

## Learning objectives 

1. Understand why feature selection is important 
2. Recognize basic strategies for identifying highly variable genes.
3. Be able to describe what is dimensionality reduction and name at least three types of methods.
4. Be able to describe what does clustering give us and name at least three types of approaches to clustering. 
5. Be able to explain why the question "What are the 'true' number of clusters?" is not a great question and be able to reframe the question to e.g. "how well do the clusters approximate the cell types or states of interest?"


## Materials 

We will go through these slides on "More Single-Cell Data Science": 

- https://docs.google.com/presentation/d/19EIrGrbZoQUyYQggy8Zxzo5RiwM7fEsiP3DSAsnCOgE/edit?usp=sharing


# Part 2

## Learning objectives 

1. Be able to load a `SingleCellExperiment` object in a `.HDF5` file format for large-scale single-cell data. 
2. Be able to apply standard scRNA-seq workflow, but to a "large" dataset. 


## Overview 

This tutorial gives a demo of large-scale clustering in the Bioconductor using the *DelayedArray* framework. 
*DelayedArray* is like an ordinary array in R, but allows for the data to be in-memory, on-disk in a file, or even hosted on a remote server. 

We will showcase an end-to-end clustering pipeline, starting from the count data matrix stored in HDF5 (similar to what one would download from the HCA data portal) all the way to visualization and interpretation of the clustering results.

1. `scran` normalization + PCA
2. `glmpca` with `scry`

Using these reduced dimensions, we present two types of clustering: 

1. SNN + Louvain clustering
2. mini-batch _k_-means (`mbkmeans`)

While we use a small-ish dataset for this demo for convenience, the code is computationally efficient even for (very) large datasets.

```{r packages}
suppressMessages({
  library(SingleCellExperiment)
  library(TENxPBMCData)
  library(scater)
  library(scran)
  library(scry)
  library(mbkmeans)
})
```

## Getting the data

```{r data}
sce <- TENxPBMCData("pbmc4k")
sce
counts(sce)
seed(counts(sce))
```

In this tutorial, we use a small dataset for the sake of running all the code in a short amount of time. However, this workflow is designed for large data and it will run just fine with any sized dataset. For instance, we have analyzed 1.3 million cells on a machine with a moderately sized RAM (e.g., 64GB).

By running the code below, you will run the workflow on the [10X Genomics 1.3 million cells dataset. (Warning: it takes some time!) Alternatively, you can substitute the code below with your own data in `SingleCellExperiment` format.

```{r million, eval=FALSE}
library(TENxBrainData)
sce <- TENxBrainData()
sce
```

## Filtering and normalization

### Removing low-quality cells

First, we use the `scater` package to compute a set of 
QC measures and filter out the low-quality samples.

Here, we exclude those cells that have a too high percentage of mitochondrial genes or for which we detect too few genes.

```{r filter}
sce <- addPerCellQC(sce, 
            subsets = list(Mito = grep("^MT-", rowData(sce)$Symbol_TENx)))
high_mito <- isOutlier(sce$subsets_Mito_percent, 
                       nmads = 3, type="higher")
low_detection <- (sce$detected < 1000)
high_counts <- sce$sum > 45000
sce <- sce[,!high_mito & !low_detection & !high_counts]
sce
```

### Removing lowly expressed genes

Next, we remove the lowly expressed genes. Here, 
we keep only those genes that have at least 1 UMI 
in at least 5% of the data. These threshold are
dataset-specific and may need to be taylored to 
specific applications.

```{r qc-gene-filter}
num_reads <- 1
num_cells <- 0.01*ncol(sce)
keep <- which(DelayedArray::rowSums(counts(sce) >= num_reads ) >= num_cells)
sce <- sce[keep,]
sce
```

These leaves us with `length(keep)` genes.

### Normalization

Here, we apply `mbkmeans` (`k=10` and batch size of 500) as a preliminary step to `scran` normalization.

```{r mbkmeans_full}
set.seed(19)
mbk <- mbkmeans(sce, whichAssay = "counts", reduceMethod = NA,
                  clusters=10, batch_size = 500)
sce$mbk10 <- paste0("mbk", mbk$Clusters)
table(mbk$Clusters)
```

We then compute the normalization factors and normalize the data.

```{r scran}
sce <- computeSumFactors(sce, cluster=mbk$Clusters, min.mean = 0.1)
sce <- logNormCounts(sce)
sce
```

## Dimensionality reduction

### PCA on normalized values

Here, we compute the first 50 principal components using the top variable genes.

```{r pca}
sce <- scater::runPCA(sce, ncomponents = 50,
                      ntop = 1000,
                      scale = TRUE,
                      BSPARAM = BiocSingular::RandomParam())
plotPCA(sce, colour_by = "mbk10")
```

### GLM-PCA

An alternative to PCA on normalized data is to use the GLM-PCA approach, implemented in the `scry` Bioconductor package. Here, we use the faster, approximate approach that computes the null residuals and runs PCA on them.

Other approaches implemented in Bioconductor for dimensionality reduction include correspondence analysis (in the `corral` package) and ZINB-WaVE (in the `zinbwave` and `NewWave` packages).

```{r glmpca}
sce <- nullResiduals(sce, assay="counts", type="deviance")
sce <- scater::runPCA(sce, ncomponents = 50,
                      ntop = 1000,
                      exprs_values = "binomial_deviance_residuals",
                      scale = TRUE, name = "GLM-PCA",
                      BSPARAM = BiocSingular::RandomParam())
plotReducedDim(sce, dimred = "GLM-PCA", colour_by = "mbk10")
```

## Clustering

Here, we use the GLM-PCA results to obtain the final cluster labels. We use two alternative approaches: Louvain and mini-batch _k_-means.

### Louvain

```{r louvain}
g <- buildSNNGraph(sce, k=10, use.dimred = "GLM-PCA")
lou <- igraph::cluster_louvain(g)
sce$louvain <- paste0("Louvain", lou$membership)
table(sce$louvain)
```

If you want more control on the resolution of the clustering, you can use the Louvain implementation available in the `resolution` package. Alternatively, the `leiden` package implements the Leiden algorithm.

### Mini-batch k-means

Mini-batch $k$-means is a faster version of $k$-means that uses only a random "mini-batch" of data at each iteration. The algorithm is fast enough to cluster the 1.3 million cell data in the space of the top 50 PC in under 30 seconds.

Here, we run it multiple times to select the value of $k$ with the elbow method.

```{r select_k}
k_list <- seq(5, 20)
km_res <- lapply(k_list, function(k) {
    mbkmeans(sce, clusters = k, 
             batch_size = 500,
             reduceMethod = "GLM-PCA",
             calc_wcss = TRUE)
})
wcss <- sapply(km_res, function(x) sum(x$WCSS_per_cluster))
plot(k_list, wcss, type = "b")
```

```{r minibatch}
sce$kmeans <- paste0("mbk", km_res[[which(k_list==12)]]$Clusters)
table(sce$kmeans)
table(sce$kmeans, sce$louvain)
```

## Cluster visualization

We can use UMAP or t-SNE to visualize the clusters.

```{r umap}
sce <- scater::runUMAP(sce, dimred = "GLM-PCA", 
                       external_neighbors = TRUE,
                       BNPARAM = BiocNeighbors::AnnoyParam())
plotUMAP(sce, colour_by = "louvain")
plotUMAP(sce, colour_by = "kmeans")
```

```{r tsne}
sce <- scater::runTSNE(sce, dimred = "GLM-PCA", 
                       external_neighbors = TRUE,
                       BNPARAM = BiocNeighbors::AnnoyParam())
plotTSNE(sce, colour_by = "louvain")
plotTSNE(sce, colour_by = "kmeans")
```


## Session info 

```{r}
sessionInfo()
```