基于列名正则表达式与元素值过滤单个矩阵的技术问询

Since you're working with a sample-by-sequence count matrix (like an OTU table), here are practical, reproducible solutions in both R and Python—two of the most common tools for this kind of data. I’ll cover the most frequent filtering scenarios you might need:

Common Filtering Scenarios & Code #

First, let’s assume your input .txt file is tab-separated, with sequence IDs as the first column (row names) and sample names as column headers.

1. Read Your Matrix #

R (using tidyverse/base R):

# Load tidyverse for easier manipulation (optional but recommended)
library(tidyverse)

# Read the matrix—row.names=1 uses the first column as row IDs, check.names=FALSE preserves your column names
otu_table <- read.delim("your_input_matrix.txt", row.names = 1, check.names = FALSE)

Python (using pandas):

import pandas as pd

# Read tab-separated matrix, index_col=0 sets sequence IDs as row index
otu_df = pd.read_csv("your_input_matrix.txt", sep="\t", index_col=0)

2. Filter Rows (Sequences) by Minimum Total Count #

Remove sequences that don’t meet a minimum total count across all samples (e.g., keep sequences with total count ≥ 10):

R:

# Base R approach (fast for large matrices)
filtered_rows_count <- otu_table[rowSums(otu_table) >= 10, ]

# Tidyverse approach (more readable for complex operations)
filtered_rows_count <- otu_table %>% 
  rowwise() %>% 
  mutate(total_count = sum(c_across(everything()))) %>% 
  filter(total_count >= 10) %>% 
  select(-total_count)

Python:

filtered_rows_count = otu_df[otu_df.sum(axis=1) >= 10]

3. Filter Rows by Prevalence #

Remove sequences that aren’t present in enough samples (e.g., keep sequences found in at least 3 samples):

R:

# Base R
filtered_rows_prevalence <- otu_table[rowSums(otu_table > 0) >= 3, ]

# Tidyverse
filtered_rows_prevalence <- otu_table %>% 
  rowwise() %>% 
  mutate(prevalence = sum(c_across(everything()) > 0)) %>% 
  filter(prevalence >= 3) %>% 
  select(-prevalence)

Python:

filtered_rows_prevalence = otu_df[(otu_df > 0).sum(axis=1) >= 3]

4. Filter Columns (Samples) by Minimum Total Count #

Remove samples with too few total reads (e.g., keep samples with ≥ 1000 total counts):

R:

# Base R
filtered_samples <- otu_table[, colSums(otu_table) >= 1000]

# Tidyverse
filtered_samples <- otu_table %>% 
  select(where(~sum(.x) >= 1000))

Python:

filtered_samples = otu_df.loc[:, otu_df.sum(axis=0) >= 1000]

5. Keep Specific Sequences or Samples #

If you have a list of sequences/samples you want to retain explicitly:

Keep specific sequences:

R:

keep_seqs <- c("seq_001", "seq_042", "seq_123")
filtered_specific_seqs <- otu_table[rownames(otu_table) %in% keep_seqs, ]

Python:

keep_seqs = ["seq_001", "seq_042", "seq_123"]
filtered_specific_seqs = otu_df.loc[keep_seqs, :]

Keep specific samples:

R:

keep_samples <- c("sample_A", "sample_C", "sample_F")
filtered_specific_samples <- otu_table[, colnames(otu_table) %in% keep_samples]

Python:

keep_samples = ["sample_A", "sample_C", "sample_F"]
filtered_specific_samples = otu_df.loc[:, keep_samples]

6. Combine Multiple Filters #

You can chain filters together to apply multiple criteria at once. For example, keep sequences with ≥10 total counts AND present in ≥3 samples:

R:

combined_filtered <- otu_table[rowSums(otu_table) >=10 & rowSums(otu_table>0)>=3, ]

Python:

combined_filtered = otu_df[(otu_df.sum(axis=1)>=10) & ((otu_df>0).sum(axis=1)>=3)]

Save Your Filtered Matrix #

Once you’re done filtering, write the result back to a .txt file:

R:

write.delim(combined_filtered, "filtered_matrix.txt", sep="\t", row.names=TRUE, quote=FALSE)

Python:

combined_filtered.to_csv("filtered_matrix.txt", sep="\t")

Bonus: Using Phyloseq (R for Microbiomics) #

If you’re working with 16S/ITS data specifically, the phyloseq package has built-in functions for filtering:

library(phyloseq)

# Convert your matrix to a phyloseq object
ps <- phyloseq(otu_table(otu_table, taxa_are_rows=TRUE))

# Filter taxa with total count ≥10 and present in ≥3 samples
ps_filtered <- prune_taxa(taxa_sums(ps)>=10 & taxa_sums(ps>0)>=3, ps)

# Extract the filtered matrix back to a data frame
filtered_phyloseq <- as.data.frame(otu_table(ps_filtered))

内容的提问来源于stack exchange，提问作者Devon O'Rourke