Error in UseMethod("select_") : no applicable method for 'select_' applied to an object of class "NULL"

Question

I get a strange error when knitting this R Markdown into an HTML file. I think it has to do with some sort of incompatibility in the dplyr package with knitr.

UPDATE: I replaced the cbind chunk with the dplyr::bind_cols command, as someone suggested below not to use cbind with dplyr. However, I now get a different, equally incomprehensible error:

library(dplyr) counts.all <- bind_cols(count.tables[["SF10281"]], count.tables[["SF10282"]])

The error I get with this change (again, only when knitting):

Error in eval(expr, envir, enclos) : not compatible with STRSXP Calls: <Anonymous> ... withVisible -> eval -> eval -> bind_cols -> cbind_all -> .Call

---

Previous error with cbind instead of dplyr::bind_cols:

Running the chunks separately works fine, and I was able to knit fine until I added the last chunk (using select from dplyr).

This is the error I get:

Quitting from lines 75-77 (Analysis_SF10281_SF10282_Sep29_2015.Rmd) 
Error in UseMethod("select_") : 
  no applicable method for 'select_' applied to an object of class "NULL"
Calls: <Anonymous> ... withVisible -> eval -> eval -> <Anonymous> -> select_

This is the entire Rmd file:

Read-in gene count tables into a single list of data frames (one data frame per sample):

```{r}
count.files <- list.files(pattern = "^SF[0-9]+_counts.txt$")

count.tables <- lapply(count.files, read.table, header=T, row.names=1)

names(count.tables) <- gsub("\\_counts.txt", "", count.files)
```

Remove gene metadata columns:

```{r}
count.tables <- lapply(count.tables, `[`, -(1:5))
```

Rename cells (columns) to short version:

```{r}
count.tables <- lapply(count.tables, function(x) {names(x) <- gsub("X.diazlab.aaron.tophat_out.SF[0-9]+.Sample_(SF[0-9]+).[0-9]+.([A-Z][0-9]+).accepted_hits.bam", "\\1-\\2", names(x)); x})
```

Save object to file for later: {r} saveRDS(count.tables, file="gliomaRawCounts_10281_10282_10345_10360.rds")

Make a single data frame with all 4 samples (384 cells), and write to text file:

```{r}
counts.all <- cbind(count.tables[["SF10281"]], count.tables[["SF10282"]], count.tables[["SF10345"]], count.tables[["SF10360"]])

write.table(counts.all, file="gliomaRawCounts_10281_10282_10345_10360.txt", sep="\t", quote=F, col.names=NA)
```

Read metadata. Do not assign cell ID column as row.names, for compatibility with dplyr.

```{r}
meta <- read.delim("QC_metrics_SCell_SF10281_SF10282_SF10345_SF10360.txt", check.names = F, stringsAsFactors = F)
```

Filter cells based on live/dead/multi calls. Exclude empty, red-only, and multi-cell wells:

```{r, results='hide', message=FALSE, warning=FALSE}
library(dplyr)
meta.select <- filter(meta, grepl("^1g", `Live-dead_call`))
```

Filter cells based on 1,000 gene threshold:

(Includes 12 'FAIL' cells)
```{r}
meta.select <- filter(meta.select, Genes_tagged > 1000)
```

Subset counts table to include only cells that passed QC.

```{r}
counts.select <- dplyr::select(counts.all, one_of(meta.select$ID))
head(counts.select[,1:10])
```

Answer 1

I just realized how old this post it, however, I'll still add in my thoughts:

It's a little hard to follow along since we don't have access to the data. Can you try to use this as a reproducible example? I added print() to the end of each statement so you can quickly view the results.

I think some of the data isn't coming through as expected, check the results based on the data structure assumptions I made here to see where our two outputs diverge

library(tidyverse)

count.tables <- 
  list(
    SF10281 = head(mpg) |> select(SF10281 = cty),
    SF10282 = head(mpg) |> select(SF10282 = cty),
    SF10345 = head(mpg) |> select(SF10345 = cty),
    SF10360 = head(mpg) |> select(SF10360 = cty)
  ) |> 
  print()

counts.all <- # can also just use this purrr::map_dfc(count.tables, bind_cols)
  bind_cols(
    count.tables$SF10281,
    count.tables$SF10282,
    count.tables$SF10345,
    count.tables$SF10360
  ) |> 
  print()

meta <- 
  tibble(
    ID = names(counts.all),
    `Live-dead_call` = rep("1g", 4),
    Genes_tagged = c(400, 800, 1200, 1600)
  ) |> 
  janitor::clean_names() |>  # so everything is snake_case and lower case
  print()

meta.select <- 
  meta |>
  filter(
    str_detect(live_dead_call, "^1g"),
    genes_tagged > 1000
  ) |> 
  print()
  

counts.select <- 
  counts.all |> 
  select(one_of(meta.select$id)) |>
  print()

Answer 2

This is a very common error we see often when working with large, complex datasets so the question deserves a generalizable solution.

In general, error messages like this:

Error in UseMethod("select_") : no applicable method for 'select_' applied to an object of class "NULL"

tell you that one of the selection criteria is returning an empty set.

A useful trouble-shooting method is to test each of the select commands to examine what it is returning. For this example, create temporary debug variables:

tempvar1 <- count.tables[["SF10281"]]
tempvar2 <- count.tables[["SF10282"]]

I suspect both of these will be empty.

In general, these problems can have either of two sources:

1. The source data files have some corruption or whitespace - use tidyverse read_table() function which does some additional cleaning.

2. "SF10281" and "SF10282" may not be recognized names at the top level of the data structure - this can be checked by running names(count.tables) to view the list of names. If missing, the source files may be differently named or located in a different directory.

In fact, knitter assumes a different working directory - so doublecheck that knitter is reading from the subdirectory you intend. See: Setting Working Directory using knitr::opts_knit

I hope these tips help others who run into a similar error message.