如何按组条件筛选最优基因？修正data.table代码逻辑问题

Got it, let's work through this to fix your gene selection logic. The main problem with your current code is that it's not first narrowing down to only the genes within 0.05 of the group's highest score before applying the count-based filters. Let's rewrite this properly to match your exact requirements:

Corrected Code #

library(data.table)
library(dplyr) # For case_when, can replace with data.table-native logic if preferred

# Combine your original data and the problematic Group 2 data
df <- structure(list(Group = c(1L, 1L, 1L, 2L, 2L, 3L, 3L, 4L, 4L, 4L, 2L, 2L, 2L, 2L, 2L), 
                     Gene = c("AQP11", "CLNS1A", "RSF1", "CFDP1", "CHST6", "ACE", "NOS2", "Gene1","Gene2","Gene3", "CFDP1", "CHST6", "RNU6-758P", "Gene1", "TMEM170A"), 
                     Score = c(0.5566507, 0.2811747, 0.5269924, 0.4186066, 0.4295135, 0.634, 0.6345, 0.7, 0.62, 0.61, 0.551740109920502, 0.598918557167053, 0.564491391181946, 0.567291617393494, 0.616708278656006 ), 
                     direct_count = c(4L, 0L, 3L, 1L, 1L, 1L, 1L, 0L, 1L, 0L, 1, 1, 0, 0, 0), 
                     secondary_count = c(5L, 2L, 6L, 2L, 3L, 1L, 1L, 0L, 0L, 1L, 62, 6, 1, 1, 2)), 
                row.names = c(NA, -15L), class = c("data.table", "data.frame"))

setDT(df)

new_df <- df[, {
  # 1. Identify the highest score in the group
  max_score <- max(Score)
  # 2. Filter to only genes within 0.05 of the max score (candidate pool)
  candidates <- .SD[abs(Score - max_score) <= 0.05]
  
  # 3. Apply selection rules to the candidate pool
  if (nrow(candidates) == 1) {
    # Condition 1: Only one candidate (max score is >0.05 away from all others)
    selected <- candidates
  } else {
    # Condition 2: Pick candidates with highest direct_count
    max_direct <- max(candidates$direct_count)
    direct_top <- candidates[direct_count == max_direct]
    
    if (nrow(direct_top) == 1) {
      selected <- direct_top
    } else {
      # Condition 3: Tie on direct_count, pick highest secondary_count
      max_secondary <- max(direct_top$secondary_count)
      secondary_top <- direct_top[secondary_count == max_secondary]
      
      if (nrow(secondary_top) == 1) {
        selected <- secondary_top
      } else {
        # Condition 4: All counts are identical, keep all candidates
        selected <- secondary_top
      }
    }
  }
  
  # Add explanatory comments matching your expected output
  comment_text <- case_when(
    nrow(candidates) == 1 ~ "#评分差＞0.05的最高分基因",
    nrow(direct_top) == 1 ~ "#最高direct_count",
    nrow(secondary_top) == 1 ~ "#direct_count相同时选最高secondary_count",
    TRUE ~ "#所有计数均相同"
  )
  selected[, Comment := comment_text]
  selected
}, by = Group]

# View the result
print(new_df, row.names = FALSE)

What This Fixes #

• Candidate Pool First: We first narrow down the group to only genes that are within 0.05 of the highest score. This ensures we never apply count filters to genes that are too far from the top score, which was the root of your earlier error.
• Step-by-Step Rule Application:
  • Starts with checking if the top score is clearly isolated (condition 1)
  • Then prioritizes direct_count among valid candidates (condition 2)
  • Breaks ties with secondary_count (condition 3)
  • Keeps all candidates if all counts are identical (condition 4)
• Clear Annotations: Added a Comment column to match the explanatory notes in your expected output.

Testing the Problematic Group 2 #

For your tricky Group 2 where TMEM170A has the highest score (0.6167), the candidate pool includes genes with scores ≥ 0.6167 - 0.05 = 0.5667:

• CHST6 (0.5989), Gene1 (0.5673), TMEM170A (0.6167)
  Among these, CHST6 has the highest direct_count (1 vs. 0 for the others), so it's correctly selected—fixing the earlier issue where your code picked Gene1 incorrectly.

内容的提问来源于stack exchange，提问作者DN1