String Ops

Strings are where bioinformatics data is messiest. Gene symbols, sample IDs, chromosome labels: all text, all inconsistent across files. stringr in R and the .str namespace in Polars cover the same operations. The one that saves the most time is the last: cleaning sample names so a join stops silently failing.

Every code block runs in a container in the companion repository, and the two languages agree on the results.

Basics: length, extract, detect

Measure a string, pull a number out of it, and test for a pattern. Here, the chromosome number is extracted from labels like chr17 by removing the chr prefix.

R

library(stringr)

annotation |>
  mutate(
    gene_len  = str_length(gene),
    chrom_num = as.integer(str_remove(chromosome, "chr")),
    is_chr17  = str_detect(chromosome, "chr17")
  )

   gene          pathway chromosome gene_len chrom_num is_chr17
1  TP53        Apoptosis      chr17        4        17     TRUE
2  EGFR Growth signaling       chr7        4         7    FALSE
3   MYC Growth signaling       chr8        3         8    FALSE
4 BRCA1       DNA repair      chr17        5        17     TRUE
5 GAPDH     Housekeeping      chr12        5        12    FALSE
6  ACTB     Housekeeping       chr7        4         7    FALSE
7  KRAS Growth signaling      chr12        4        12    FALSE

Python

annotation.with_columns(
    pl.col("gene").str.len_chars().alias("gene_len"),
    pl.col("chromosome").str.replace("chr", "").cast(pl.Int64).alias("chrom_num"),
    pl.col("chromosome").str.contains("chr17").alias("is_chr17"),
)

shape: (7, 6)
┌───────┬──────────────────┬────────────┬──────────┬───────────┬──────────┐
│ gene  ┆ pathway          ┆ chromosome ┆ gene_len ┆ chrom_num ┆ is_chr17 │
│ ---   ┆ ---              ┆ ---        ┆ ---      ┆ ---       ┆ ---      │
│ str   ┆ str              ┆ str        ┆ u32      ┆ i64       ┆ bool     │
╞═══════╪══════════════════╪════════════╪══════════╪═══════════╪══════════╡
│ TP53  ┆ Apoptosis        ┆ chr17      ┆ 4        ┆ 17        ┆ true     │
│ EGFR  ┆ Growth signaling ┆ chr7       ┆ 4        ┆ 7         ┆ false    │
│ MYC   ┆ Growth signaling ┆ chr8       ┆ 3        ┆ 8         ┆ false    │
│ BRCA1 ┆ DNA repair       ┆ chr17      ┆ 5        ┆ 17        ┆ true     │
│ GAPDH ┆ Housekeeping     ┆ chr12      ┆ 5        ┆ 12        ┆ false    │
│ ACTB  ┆ Housekeeping     ┆ chr7       ┆ 4        ┆ 7         ┆ false    │
│ KRAS  ┆ Growth signaling ┆ chr12      ┆ 4        ┆ 12        ┆ false    │
└───────┴──────────────────┴────────────┴──────────┴───────────┴──────────┘

Cleaning messy sample names

This is the one that bites everyone. The same sample arrives labelled s01, S_02, S-03 , s04: mixed case, a hyphen where another file used an underscore, a stray leading or trailing space. None of these match S01 on a join, and nothing warns you. The fix is three steps: strip whitespace, upcase, remove the separators.

R

messy <- read.csv("../fixtures/messy_samples.csv")
messy |>
  mutate(sample_id = str_replace_all(str_to_upper(str_trim(sample_label)), "[-_ ]", ""))

  sample_label rin sample_id
1          s01 9.1       S01
2         S_02 8.7       S02
3        S-03  9.4       S03
4          s04 8.2       S04
5          S05 9.0       S05
6         S 06 7.9       S06
7         s-07 8.8       S07
8         S08  9.3       S08

Python

messy = pl.read_csv("../fixtures/messy_samples.csv")
messy.with_columns(
    pl.col("sample_label")
    .str.strip_chars()
    .str.to_uppercase()
    .str.replace_all("[-_ ]", "")
    .alias("sample_id")
)

shape: (8, 3)
┌──────────────┬─────┬───────────┐
│ sample_label ┆ rin ┆ sample_id │
│ ---          ┆ --- ┆ ---       │
│ str          ┆ f64 ┆ str       │
╞══════════════╪═════╪═══════════╡
│ s01          ┆ 9.1 ┆ S01       │
│ S_02         ┆ 8.7 ┆ S02       │
│ S-03         ┆ 9.4 ┆ S03       │
│  s04         ┆ 8.2 ┆ S04       │
│ S05          ┆ 9.0 ┆ S05       │
│ S 06         ┆ 7.9 ┆ S06       │
│ s-07         ┆ 8.8 ┆ S07       │
│ S08          ┆ 9.3 ┆ S08       │
└──────────────┴─────┴───────────┘

Why it matters: the join

The payoff is concrete. Join the sample labels to a per-sample summary before cleaning, and only the one already-canonical label (S05) matches. After cleaning, all eight do.

R

raw_match   <- inner_join(messy, sample_totals, by = c("sample_label" = "sample_id"))
clean_match <- inner_join(messy, sample_totals, by = "sample_id")

raw labels matched:     1 of 8
cleaned labels matched: 8 of 8

Python

raw_match = messy.join(sample_totals, left_on="sample_label", right_on="sample_id", how="inner")
clean_match = messy.join(sample_totals, on="sample_id", how="inner")

raw labels matched:     1 of 8
cleaned labels matched: 8 of 8

Seven samples silently dropped, with no error, until you clean the key. When a join returns fewer rows than you expect, the labels are the first thing to check.