Core Verbs

Most data wrangling is a handful of verbs applied over and over: keep some rows, keep some columns, sort, and compute new columns. dplyr in R and Polars in Python express the same four ideas with different words. This page runs both on one tidy bioinformatics table, so you can read a verb in whichever language you know and see its twin in the other.

Every code block runs in a container in the companion repository. The tables shown are the real output of those runs, and the two languages agree on every number.

The verbs at a glance

Polars borrowed some dplyr names, so two of the four verbs are the same word in both. The other two differ.

Operation	dplyr (R)	Polars (Python)
keep rows	filter	filter
keep columns	select	select
sort rows	arrange	sort
add a column	mutate	with_columns

A shared name is not a shared syntax. dplyr reads a column as if it were a variable in scope (filter(expr, condition == "treated")), while Polars builds an expression on an explicit column object (expr.filter(pl.col("condition") == "treated")), and combines conditions with & and | rather than commas. That pl.col(...) expression style is the real difference between the two, more than the verb names.

The data

expression.csv is a long counts table: one row per sample and gene. Eight samples, four control and four treated, measured across six genes.

R

library(dplyr)

expr <- read.csv("../data/expression.csv")
head(expr)

  sample_id condition  gene counts
1       S01   control  TP53    393
2       S01   control  EGFR    294
3       S01   control   MYC    247
4       S01   control BRCA1    195
5       S01   control GAPDH   3939
6       S01   control  ACTB   3117

Python

import polars as pl

expr = pl.read_csv("../data/expression.csv")
expr.head()

shape: (5, 4)
┌───────────┬───────────┬───────┬────────┐
│ sample_id ┆ condition ┆ gene  ┆ counts │
│ ---       ┆ ---       ┆ ---   ┆ ---    │
│ str       ┆ str       ┆ str   ┆ i64    │
╞═══════════╪═══════════╪═══════╪════════╡
│ S01       ┆ control   ┆ TP53  ┆ 393    │
│ S01       ┆ control   ┆ EGFR  ┆ 294    │
│ S01       ┆ control   ┆ MYC   ┆ 247    │
│ S01       ┆ control   ┆ BRCA1 ┆ 195    │
│ S01       ┆ control   ┆ GAPDH ┆ 3939   │
└───────────┴───────────┴───────┴────────┘

filter: keep the rows you want

Subset rows by a condition. Here, only the treated samples. dplyr names the column directly; Polars wraps it in pl.col.

R

treated <- filter(expr, condition == "treated")
head(treated)

  sample_id condition  gene counts
1       S05   treated  TP53    619
2       S05   treated  EGFR    757
3       S05   treated   MYC    810
4       S05   treated BRCA1    204
5       S05   treated GAPDH   4376
6       S05   treated  ACTB   3292

Python

treated = expr.filter(pl.col("condition") == "treated")
treated.head()

shape: (5, 4)
┌───────────┬───────────┬───────┬────────┐
│ sample_id ┆ condition ┆ gene  ┆ counts │
│ ---       ┆ ---       ┆ ---   ┆ ---    │
│ str       ┆ str       ┆ str   ┆ i64    │
╞═══════════╪═══════════╪═══════╪════════╡
│ S05       ┆ treated   ┆ TP53  ┆ 619    │
│ S05       ┆ treated   ┆ EGFR  ┆ 757    │
│ S05       ┆ treated   ┆ MYC   ┆ 810    │
│ S05       ┆ treated   ┆ BRCA1 ┆ 204    │
│ S05       ┆ treated   ┆ GAPDH ┆ 4376   │
└───────────┴───────────┴───────┴────────┘

select: keep the columns you want

Pick a subset of columns by name.

R

select(expr, sample_id, gene, counts)

  sample_id  gene counts
1       S01  TP53    393
2       S01  EGFR    294
3       S01   MYC    247
4       S01 BRCA1    195
5       S01 GAPDH   3939
6       S01  ACTB   3117

Python

expr.select("sample_id", "gene", "counts")

shape: (5, 3)
┌───────────┬───────┬────────┐
│ sample_id ┆ gene  ┆ counts │
│ ---       ┆ ---   ┆ ---    │
│ str       ┆ str   ┆ i64    │
╞═══════════╪═══════╪════════╡
│ S01       ┆ TP53  ┆ 393    │
│ S01       ┆ EGFR  ┆ 294    │
│ S01       ┆ MYC   ┆ 247    │
│ S01       ┆ BRCA1 ┆ 195    │
│ S01       ┆ GAPDH ┆ 3939   │
└───────────┴───────┴────────┘

arrange / sort: order the rows

Sort by a column, highest first. The verb is arrange(desc(...)) in dplyr and sort(..., descending=True) in Polars. The housekeeping genes ACTB and GAPDH top the list, as expected.

R

arrange(expr, desc(counts))

  sample_id condition  gene counts
1       S03   control  ACTB   4398
2       S05   treated GAPDH   4376
3       S02   control  ACTB   4331
4       S08   treated GAPDH   4312
5       S06   treated  ACTB   4199
6       S03   control GAPDH   4118

Python

expr.sort("counts", descending=True)

shape: (5, 4)
┌───────────┬───────────┬───────┬────────┐
│ sample_id ┆ condition ┆ gene  ┆ counts │
│ ---       ┆ ---       ┆ ---   ┆ ---    │
│ str       ┆ str       ┆ str   ┆ i64    │
╞═══════════╪═══════════╪═══════╪════════╡
│ S03       ┆ control   ┆ ACTB  ┆ 4398   │
│ S05       ┆ treated   ┆ GAPDH ┆ 4376   │
│ S02       ┆ control   ┆ ACTB  ┆ 4331   │
│ S08       ┆ treated   ┆ GAPDH ┆ 4312   │
│ S06       ┆ treated   ┆ ACTB  ┆ 4199   │
└───────────┴───────────┴───────┴────────┘

mutate / with_columns: add a derived column

Compute a new column from existing ones. Raw counts are usually log-transformed before analysis, so here we add log2(counts + 1). dplyr calls it mutate; Polars calls it with_columns and needs an explicit .alias for the new name.

R

expr <- mutate(expr, log2_expr = log2(counts + 1))
head(expr)

  sample_id condition  gene counts log2_expr
1       S01   control  TP53    393  8.622052
2       S01   control  EGFR    294  8.204571
3       S01   control   MYC    247  7.954196
4       S01   control BRCA1    195  7.614710
5       S01   control GAPDH   3939 11.943980
6       S01   control  ACTB   3117 11.606405

Python

expr = expr.with_columns((pl.col("counts") + 1).log(2).alias("log2_expr"))
expr.head()

shape: (5, 5)
┌───────────┬───────────┬───────┬────────┬───────────┐
│ sample_id ┆ condition ┆ gene  ┆ counts ┆ log2_expr │
│ ---       ┆ ---       ┆ ---   ┆ ---    ┆ ---       │
│ str       ┆ str       ┆ str   ┆ i64    ┆ f64       │
╞═══════════╪═══════════╪═══════╪════════╪═══════════╡
│ S01       ┆ control   ┆ TP53  ┆ 393    ┆ 8.622052  │
│ S01       ┆ control   ┆ EGFR  ┆ 294    ┆ 8.204571  │
│ S01       ┆ control   ┆ MYC   ┆ 247    ┆ 7.954196  │
│ S01       ┆ control   ┆ BRCA1 ┆ 195    ┆ 7.61471   │
│ S01       ┆ control   ┆ GAPDH ┆ 3939   ┆ 11.94398  │
└───────────┴───────────┴───────┴────────┴───────────┘

Chaining the verbs

The verbs compose. dplyr threads a table through the native pipe |>; Polars chains methods on the frame. Both read the same way: take the treated MYC rows, keep three columns, sort by expression.

R

expr |>
  filter(gene == "MYC", condition == "treated") |>
  select(sample_id, counts, log2_expr) |>
  arrange(desc(log2_expr))

  sample_id counts log2_expr
1       S07    854  9.739781
2       S08    834  9.705632
3       S05    810  9.663558
4       S06    531  9.055282

Python

(
    expr.filter((pl.col("gene") == "MYC") & (pl.col("condition") == "treated"))
    .select("sample_id", "counts", "log2_expr")
    .sort("log2_expr", descending=True)
)

shape: (4, 3)
┌───────────┬────────┬───────────┐
│ sample_id ┆ counts ┆ log2_expr │
│ ---       ┆ ---    ┆ ---       │
│ str       ┆ i64    ┆ f64       │
╞═══════════╪════════╪═══════════╡
│ S07       ┆ 854    ┆ 9.739781  │
│ S08       ┆ 834    ┆ 9.705632  │
│ S05       ┆ 810    ┆ 9.663558  │
│ S06       ┆ 531    ┆ 9.055282  │
└───────────┴────────┴───────────┘

Same rows, same order, same values. Across the full table the two agree exactly: 24 treated rows, 76872 total counts, and a mean log2 expression of 9.541 for MYC in the treated group. That agreement is checked on every run by the harness parity gate, so this page cannot drift from its code.