Skip to content

Commit

Permalink
template: general review and fixes
Browse files Browse the repository at this point in the history
  • Loading branch information
naikymen committed Jun 3, 2024
1 parent 085f342 commit dde1294
Showing 1 changed file with 87 additions and 84 deletions.
171 changes: 87 additions & 84 deletions inst/rmarkdown/templates/rmd_template.dev/skeleton/skeleton.Rmd
View file
Original file line number Diff line number Diff line change
Expand Up @@ -230,7 +230,7 @@ Load "pdata":

```{r}
# Set the path to the file (modify accodringly).
pdata_path <- "/.../somewhere/over/the/pdata.csv"
pdata_path <- ".../somewhere/over/the/pdata.csv"
# It is a "metadata" dataframe we usually join to cdata by "position" and/or "t.frame",
# which contains other experimental variables: simuli concentrations, notes, etc.
Expand All @@ -251,24 +251,6 @@ cdata <- left_join(cdata, pdata)

<!-- - [ ] TO-DO -->

### Shiny-Magick framework

Extra tools for graphical filtering and annotation/tagging are available in
the `rcell2.magick` package.

That package also includes functions for generating and manipulating
images of single cells, and preparing strips, tiles, and plots.

To learn more, install the `rcell2.magick` package and open it's template notebook.

Uncomment and execute the following to install the package and open the template:

```{r}
# if(!requireNamespace("rcell2.magick")) remotes::install_github("darksideoftheshmoo/rcell2-magick")
# rcell2.magick::get_workflow_template_magick()
```

### Cell-tagging using ImageJ/FIJI

No analysis package exists (yet) that can replace a researcher's general visual assesment.
Expand Down Expand Up @@ -325,58 +307,25 @@ Steps (continued): read these steps before tagging.
specially relevant for `t.frame`s.
6. After tagging the positions make measurements (press Ctrl+M) and save
the output to a `.csv` file.
7. Prepare a `data.frame` with tag names mapped to _Counter_ indices.
- Note that ImageJ's Multi-point counters begin at 0.

Each "Counter" has an associated "tag", corresponding to the event or class
marked in the images.

```{r}
tag_descriptions <- c("Ok", "Budding", "Dead", "Spurious")
7. Load the Multi-point tags.

counter.desc.df <- data.frame(
Counter = 0:3,
tag_description = tag_descriptions
)
counter.desc.df
```

8. Prepare a `data.frame` relating _Slice_ indices to stage position indices.
- Note that ImageJ's _Slices_ begin at 1.

```{r}
slice.pos.df <- data.frame(
Slice = 1:2,
pos = c(4,12)
)
slice.pos.df
```
We usually prepare one CSV file per position, and extract the position index
from the file name when loading the measurements into R.
For example, we CSV files saved as `pos01_filter.csv` (for position 1).

9. Prepare a `data.frame` relating _Ch_ indices to `t.frame` position indices.
- Note that ImageJ's _Ch_ begin at 1, and Cell-ID's `t.frame`s begin at 0.
This will require a regex with a single capturing group for the position index, such as:

```{r}
ch.frame.df <- data.frame(
Ch = 1:3,
t.frame = 0:2
)
ch.frame.df
results.pattern <- "pos(\\d+)_filter.csv"
```

9. Load the Multi-point tags.

We usually prepare one CSV file per position, and extract the position index
from the file name when loading the measurements into R.
For example, we CSV files saved as `pos01_filter.csv` (for position 1).

List CSV files:

```{r}
results.dir <- ".../somewhere/over/the/results/"
mp.filters.files <- dir(path = results.dir,
pattern = "pos\\d+_filter.csv",
pattern = results.pattern,
full.names = T)
mp.filters.files
Expand All @@ -392,7 +341,7 @@ mp.filters <-
# Bind rows
bind_rows(.id = "file") %>%
# Get position info
extract(file, "pos(\\d+)_", into = "pos", convert = T) %>%
extract(file, results.pattern, into = "pos", convert = T) %>%
# Cleanup
select(pos, X, Y, Ch, Slice, Frame) %>%
# Add an ID to each tag
Expand All @@ -401,21 +350,38 @@ mp.filters <-
mp.filters
```

10. Reindex the time frame column to match Cell-ID's.
8. Re-index the time frame column to match Cell-ID's.

```{r}
mp.filters <-
# Re-index t.frame
mutate(t.frame = t.frame - 1)
mutate(t.frame = Frame - 1)
```

11. Examine the result.
9. Examine the result.

```{r}
mp.filters
```

> Reminder: this is needed because virtual hyperstacks can only load images in "xyczt" order.
10. _Optional_: Prepare a `data.frame` with tag names mapped to _Counter_ indices.
- Note that ImageJ's Multi-point counters begin at 0.
- Each "Counter" has an associated "tag", corresponding to the event or class marked in the images.

```{r}
# For example:
tag_counter <- 0:3
tag_descriptions <- c("Ok", "Budding", "Dead", "Spurious")
counter.desc.df <- data.frame(
Counter = tag_counter,
tag_description = tag_descriptions
)
counter.desc.df
```

##### Sample dataset

Here is an example `mp.filters` `data.frame`. In this case the position
is encoded in the _Slice_ variable.
Expand All @@ -440,23 +406,22 @@ mp.filters <- multipoints %>%
mp.filters
```

#### Mapping points
#### Map points to cells

Keep the relevant set of markers:
Keep the relevant set of counters (a.k.a. markers). This code will work with
one marker at a time.

```{r}
mp.filter <- mp.filters |>
filter(Counter == 0) |> # Here you may choose which markers to map.
# Here you may choose which markers to map.
# filter(Counter == 0) |>
select(mpid, X, Y, pos, t.frame)
mp.filter
```

Calculate distance to closest multipoint and save it's ID:

* Consider only multipoints in bounding box.
* Use counters for rtcc (0) and cycling (1).

```{r}
# Make a copy of cdata
cdata.mp <- cdata %>%
Expand All @@ -465,9 +430,7 @@ cdata.mp <- cdata %>%
# Make a copy of the points
mp.mapped <- mp.filter
```
```{r}
# Add the columns
mp.mapped[, c("ucid", "closest.dist")] <-
apply(mp.filter, 1, function(x){
Expand All @@ -487,9 +450,9 @@ mp.mapped[, c("ucid", "closest.dist")] <-
mp.mapped
```

Fix mis-assgnments manually:
Fix mis-assgnments manually here:

> If you spot any mis-assignments further on, you can fix them here mannualy and then re-inspect.
> If you spot any mis-assignments later, you can fix them here mannualy and then re-inspect.
```{r}
# mp.mapped[mp.mapped$ucid==110047,"ucid"] <- 110754
Expand All @@ -516,10 +479,14 @@ mp.mapped.minimal |>
facet_wrap(~Counter)
```

Set a threshold distance, and save the filter:
Set a threshold distance, and apply the filter:

> This value will depend on you opticalmicroscopy setup, and the morphology of your cells.
```{r}
cdata.mp$has.close.mp <- cdata.mp$closest.dist < 10
threshold <- 11
cdata.mp$has.close.mp <- cdata.mp$closest.dist < threshold
```

Plot result:
Expand All @@ -541,11 +508,12 @@ p <- ggplot() +
data=cdata.mp %>% filter(has.close.mp)) +
facet_wrap(~pos) + coord_equal() +
ggtitle("Examine MP ilter result",
ggtitle(paste0("Examine MP filter result", " (d<", threshold, ")"),
"Black: ImageJ points; Gray: all Cell-ID cells; Red: Cell-ID cells matched to a point.") +
scale_color_manual(values=c(ImageJ="black", matched="red", CellID="grey")) +
scale_y_reverse()
scale_y_reverse() +
theme_minimal()
# plotly::ggplotly(p)
p
Expand All @@ -559,7 +527,26 @@ cdata.mp %>% filter(has.close.mp) |>
rcell2.magick::magickForKnitr()
```

#### Check for duplicated IDs
Examine images of the un-matched points:

```{r}
# Make a cdata-like data frame for magickCell.
unmatched.points <- mp.filters |>
dplyr::rename(xpos=X,ypos=Y) |>
inner_join(
mp.mapped.minimal |>
filter(closest.dist >= threshold) |>
select(-t.frame),
by = "mpid"
)
# Display some images.
unmatched.points |>
rcell2.magick::magickCell(images) |>
rcell2.magick::magickForKnitr()
```

#### Find duplicates

Generate indicator column:

Expand Down Expand Up @@ -662,9 +649,26 @@ cdata.filtered %>%
rcell2.magick::magickForKnitr()
```


Repeat these steps for the other Multi-point _Counters_.

### Shiny-Magick framework

Extra tools for graphical filtering and annotation/tagging are available in
the `rcell2.magick` package.

That package also includes functions for generating and manipulating
images of single cells, and preparing strips, tiles, and plots.

To learn more, install the `rcell2.magick` package and open it's template notebook.

Uncomment and execute the following to install the package and open the template:

```{r}
# if(!requireNamespace("rcell2.magick")) remotes::install_github("darksideoftheshmoo/rcell2-magick")
# rcell2.magick::get_workflow_template_magick()
```

### Hu Moments

The Hu Moments can be computed from the masked TIFF files, generated by Cell-ID with
Expand All @@ -684,9 +688,8 @@ rcell2:::check_tiff_mask(cell_data)
Alternatively, the Hu Moments can be computed from the TSV files (see `output_coords_to_tsv`
at the help page of `rcell2.cellid::cell2`), by using the `append_hues2`.

Read its help page for more information:
Read its help page for more information and usage:

```{r}
?append_hues2
```

0 comments on commit dde1294

Please sign in to comment.