Skip to content

Commit

Permalink
Adjust figures and fix names
Browse files Browse the repository at this point in the history
  • Loading branch information
@lazappi
lazappi committed Apr 3, 2019
1 parent 33ac14f commit fabb156
Show file tree
Hide file tree
Showing 63 changed files with 1,598 additions and 389 deletions.
78 changes: 16 additions & 62 deletions analysis/01-preprocessing.Rmd
View file
Original file line number Diff line number Diff line change
Expand Up @@ -141,7 +141,7 @@ Traditional Cell Ranger
=======================

The version of Cell Ranger we have used here selects cells using a modified
version of the `emptyDrops` method (see below) but older versions used an
version of the `EmptyDrops` method (see below) but older versions used an
alternative approach by calculating the 99th percentile of the total number
of counts in top given expected number of cells and selecting droplets that
had at least 10 percent of this many counts. For comparison purposes we will
Expand Down Expand Up @@ -176,7 +176,7 @@ cells.
Empty drops
===========

We will now look at identifying which droplets to select using the `emptyDrops`
We will now look at identifying which droplets to select using the `EmptyDrops`
method. This method tests whether the composition of a droplet is significantly
different from the ambient RNA in the sample which is obtained by pooling the
empty droplets. Droplets with very large counts are also automatically retained.
Expand All @@ -190,7 +190,7 @@ emp_drops <- emptyDrops(counts(raw), lower = empty_thresh, niters = emp_iters,
test.ambient = TRUE, BPPARAM = bpparam)
```

`emptyDrops` calculates p-values using a permutation approach. Let's check that
`EmptyDrops` calculates p-values using a permutation approach. Let's check that
we are usually a sufficient number of iterations. If there are any droplets
that have non-significant p-values but are limited by the number of permuations
the number should be increased.
Expand All @@ -209,7 +209,7 @@ colData(raw)$EmpDropsFilt <- is_cell
table(Limited = emp_drops$Limited, Significant = is_cell)
```

Another way to check the `emptyDrops` results is to look at the droplets below
Another way to check the `EmptyDrops` results is to look at the droplets below
our empty threshold. We are assuming that these droplets only contain ambient
RNA and therefore the null hypothesis should be true and the distribution of
p-values should be approximately uniform.
Expand All @@ -229,7 +229,7 @@ Peaks near zero would tell us that not all of the droplets below the threshold
are truly empty and that we should lower it.

We can also plot the negative log-probability against the total counts to see
which droplets `emptyDrops` has selected.
which droplets `EmptyDrops` has selected.

```{r empty-drops-plot}
colData(raw) %>%
Expand Down Expand Up @@ -259,26 +259,26 @@ plot:
plot_data <- list(
"Cell Ranger" = colData(raw)$Cell[colData(raw)$DefaultFilt],
"Cell Ranger v3" = colData(raw)$Cell[colData(raw)$CellRangerFilt],
"emptyDrops" = colData(raw)$Cell[colData(raw)$EmpDropsFilt]
"EmptyDrops" = colData(raw)$Cell[colData(raw)$EmpDropsFilt]
)
upset(fromList(plot_data), order.by = "freq",
sets.x.label = "Number of cells", text.scale = c(2, 1.2, 2, 1.2, 2, 3))
```

We can see that most of the cells are identified by all methods. A large number
of cells are also identified by the new Cell Ranger and `emptyDrops` methods.
of cells are also identified by the new Cell Ranger and `EmptyDrops` methods.
These are likely to be those cells that fall below the total counts threshold
selected by the traditional Cell Ranger algorithm. Our use of the `emptyDrops`
selected by the traditional Cell Ranger algorithm. Our use of the `EmptyDrops`
algorithm has identified even more cells than Cell Ranger v3 but there are
very few cells that Cell Ranger identifed which `emptyDrops` didn't.
very few cells that Cell Ranger identifed which `EmptyDrops` didn't.

Selection
=========

We are going to perform further quality control of these cells anyway so at
this stage we will keep those that were selected by either Cell Ranger v3 or
`emptyDrops`.
`EmptyDrops`.

```{r select}
selected <- raw[, colData(raw)$CellRangerFilt | colData(raw)$EmpDropsFilt]
Expand Down Expand Up @@ -316,25 +316,6 @@ Figure
======

```{r figure}
# plot_data <- colData(raw) %>%
# as.data.frame() %>%
# select(Cell, Sample, Total = BarcodeTotal, Kept = DefaultFilt) %>%
# group_by(Sample) %>%
# mutate(Rank = rank(-Total)) %>%
# arrange(Sample, Rank)
#
# default_plot <- ggplot(plot_data, aes(x = Rank, y = Total)) +
# geom_point(shape = 1, aes(colour = Kept)) +
# scale_x_log10(labels = scales::number) +
# scale_y_log10(labels = scales::number) +
# scale_colour_manual(values = c("black", "#8DC63F"),
# labels = c("Empty", "Cell")) +
# facet_wrap(~ Sample, nrow = 1, labeller = label_both) +
# ylab("Total counts") +
# ggtitle("Cell Ranger") +
# theme_minimal() +
# theme(legend.title = element_blank())
plot_data <- colData(raw) %>%
as.data.frame() %>%
select(Cell, Kept = EmpDropsFilt, Rank = BarcodeRank,
Expand Down Expand Up @@ -362,7 +343,7 @@ emp_plot <- plot_data %>%
scale_y_log10(labels = scales::number, breaks = 10 ^ seq(0, 5)) +
scale_colour_manual(values = c("black", "#EC008C"),
labels = c("Empty", "Cell")) +
ggtitle("emptyDrops droplet selection") +
ggtitle("EmptyDrops droplet selection") +
ylab("Total counts") +
theme_minimal() +
theme(legend.position = "bottom",
Expand All @@ -375,32 +356,6 @@ ggsave(here::here("output", DOCNAME, "droplet-selection.png"), emp_plot,
width = 7, height = 5, scale = 1)
emp_plot
# cr_plot <- bc_data %>%
# filter(Total > 0) %>%
# ggplot(aes(x = Rank, y = Total)) +
# geom_point(shape = 1, aes(colour = Kept)) +
# scale_x_log10(labels = scales::number, breaks = 10 ^ seq(0, 6)) +
# scale_y_log10(labels = scales::number, breaks = 10 ^ seq(0, 5)) +
# scale_colour_manual(values = c("black", "#F47920"),
# labels = c("Empty", "Cell")) +
# ggtitle("Cell Ranger v3") +
# ylab("Total counts") +
# theme_minimal() +
# theme(legend.position = "bottom",
# legend.title = element_blank())
#
# p1 <- plot_grid(emp_plot, cr_plot, nrow = 1, labels = c("B", "C"))
#
# fig <- plot_grid(default_plot, p1, ncol = 1, rel_heights = c(0.5, 1),
# labels = c("A", ""))
#
# ggsave(here::here("output", DOCNAME, "droplet-selection.pdf"), fig,
# width = 7, height = 6, scale = 2)
# ggsave(here::here("output", DOCNAME, "droplet-selection.png"), fig,
# width = 7, height = 6, scale = 2)
#
# fig
```

```{r fig-comp, collapse = TRUE, fig.show = "hide"}
Expand All @@ -409,15 +364,14 @@ plot_data <- colData(raw) %>%
select(Name = Cell,
`Cell Ranger` = DefaultFilt,
`Cell Ranger v3` = CellRangerFilt,
emptyDrops = `EmpDropsFilt`,
EmptyDrops = `EmpDropsFilt`,
`Total counts` = BarcodeTotal) %>%
mutate(`Cell Ranger` = if_else(`Cell Ranger`, 1L, 0L),
`Cell Ranger v3` = if_else(`Cell Ranger v3`, 1L, 0L),
emptyDrops = if_else(emptyDrops, 1L, 0L)) %>%
EmptyDrops = if_else(EmptyDrops, 1L, 0L)) %>%
mutate(`Total counts` = log10(`Total counts`))
upset(plot_data, order.by = "freq",
#boxplot.summary = "Total counts",
sets.x.label = "Number of cells",
text.scale = c(2, 1.6, 2, 1.3, 2, 3),
matrix.color = "#7A52C7",
Expand Down Expand Up @@ -455,12 +409,12 @@ params <- list(
list(
Parameter = "emp_iters",
Value = emp_iters,
Description = "Number of iterations for emptyDrops p-values"
Description = "Number of iterations for EmptyDrops p-values"
),
list(
Parameter = "emp_fdr",
Value = emp_fdr,
Description = "FDR cutoff for emptyDrops"
Description = "FDR cutoff for EmptyDrops"
),
list(
Parameter = "n_default",
Expand All @@ -475,7 +429,7 @@ params <- list(
list(
Parameter = "n_empdrops",
Value = sum(colData(raw)$EmpDropsFilt),
Description = "Number of cells selected by the emptyDrops method"
Description = "Number of cells selected by the EmptyDrops method"
),
list(
Parameter = "n_cells",
Expand Down
Loading

0 comments on commit fabb156

Please sign in to comment.