Add native CRAM output support (unsorted and coordinate-sorted) via HTSlib

[BenjaminDEMAILLE]

Add native CRAM output support to STAR

Overview

This contribution adds full support for CRAM output in STAR, both for unsorted and coordinate-sorted alignments, using HTSlib’s native APIs. The implementation is fully backward compatible and preserves all existing BAM/SAM functionality.

---

Motivation

• CRAM is a modern, highly compressed format for storing sequencing alignments, supported by the GA4GH and major genomics tools.
• STAR previously only supported SAM and BAM output. Adding CRAM output allows users to save disk space and integrate more easily with modern pipelines.

---

Features

• New output format:
  Users can now specify --outSAMtype CRAM Unsorted or --outSAMtype CRAM SortedByCoordinate to produce CRAM files directly from STAR.

• HTSlib integration:
  Output is handled via HTSlib’s htsFile and sam_write1 APIs, ensuring compatibility with the CRAM standard and future-proofing for other formats.

• Automatic format detection:
  The output format is detected from the --outSAMtype parameter and the correct file mode is used for htslib.

• Unsorted and sorted output:

  • Unsorted CRAM output is written directly during alignment.
  • For coordinate-sorted output, STAR merges the sorted BAM bins and writes the result as CRAM using sam_write1.

• Backward compatibility:
  Existing BAM/SAM output logic is preserved. If CRAM is not requested, STAR behaves as before.

---

Implementation Details

• The BAMoutput class now supports both legacy BGZF output and htslib-based output (htsFile), with logic to route records to the appropriate backend.
• Output file opening in ReadAlignChunk.cpp and bamSortByCoordinate.cpp is updated to use hts_open with the correct mode for CRAM/SAM/BAM.
• The merging logic for coordinate-sorted output is updated:
  • If CRAM (or SAM/BAM via htslib) is requested, each sorted bin is read and records are written to the final output using sam_write1.
  • If BAM is requested and htslib is not used, the legacy bam_cat logic is preserved.
• All changes are guarded to ensure no impact on users who do not request CRAM output.

---

Usage

• Unsorted CRAM output:

  --outSAMtype CRAM Unsorted