--ERCC-correlation and --SIRV-concentration raise false "no data" errors with standard gene-level SIRV/ERCC IDs

[kayihui]

Summary

Both ERCCcorrelation and SIRVsuiteConcentration call countReader.read_counting_file() without specifying counting_type, so it silently falls back to the "transcript" default. The transcript-mode regex patterns in countReader.py don't match standard gene-level spike-in IDs (e.g. ERCC-00002, SIRV1, SIRV4001), so every spike-in row gets skipped during parsing. This produces misleading downstream errors that look like a data problem when it's actually a parsing problem:

Exception: Error: sample 'L1' does not contain any ERCCs! Please check if you counts file has a corresponding entry for ERCCs greater than 0!

Exception: Error: input data for sampe group 'L1' is empty! Please make sure that input count files contain some values greater than 0 for SIRV transcripts!

Root cause

In SIRVsuite/Pipeline/countReader.py:

self.spike_in_name_pattern["SIRV"]["transcript"] = "^SIRV\\d{3}$"
self.spike_in_name_pattern["ERCC"]["transcript"] = "(DQ|EF)\\d{6}"

The ERCC transcript pattern expects GenBank accession-style IDs (DQ123456), not the standard ERCC-XXXXX gene ID format that most quantification tools (Salmon, RSEM, Cufflinks, etc.) actually output. The SIRV transcript pattern only matches exactly 3-digit IDs, missing both the 1-digit gene-level IDs (SIRV1–SIRV7) and the variable-length isoform IDs (SIRV4001, SIRV10001, SIRV12001, etc.) that real SIRV reference sets contain.

Both ERCCcorrelation.__init__ (ERCC_correlation.py) and SIRVsuiteConcentration.__init__ (SIRV_concentration.py) call read_counting_file() without ever passing counting_type, so there's no way to reach the (correctly-matching) "gene" patterns through the existing CLI — the "transcript" default is hit unconditionally regardless of the actual ID format in the user's counts file.

Reproduction

import re
# ERCC: real counts file has "ERCC-00002", "ERCC-00012", etc.
re.match(r"(DQ|EF)\d{6}", "ERCC-00002")  # None — silently skipped

# SIRV: real counts file has "SIRV1"..."SIRV7" and "SIRV4001", "SIRV10001", etc.
re.match(r"^SIRV\d{3}$", "SIRV1")     # None
re.match(r"^SIRV\d{3}$", "SIRV4001")  # None

Confirmed end-to-end: a counts file with valid, non-zero ERCC/SIRV values (verified by manual inspection) reproducibly triggers both exceptions above, purely from the regex mismatch — not from any actual absence of spike-in data.

Suggested fix

self.spike_in_name_pattern["SIRV"]["transcript"] = "^SIRV\\d+$"
self.spike_in_name_pattern["ERCC"]["transcript"] = "^ERCC-\\d{5}$"

^SIRV\d+$ covers both the 1-digit gene IDs and the longer isoform IDs without needing a fixed digit count, and correctly excludes ERCC IDs and malformed strings. ^ERCC-\d{5}$ matches the standard ERCC gene ID format directly.

This is a one-line-per-pattern change and resolves both reported exceptions in local testing.

---

Diagnosed and verified with help from Claude (Anthropic).

---