Add configurable --display-thresh parameter for .prob file filtering

[Copilot]

Rebased PR - Display Thresh Parameter Only

Completed Changes:

• Reset branch to main (af64207)
• Applied only display-thresh parameter changes
• Verified no debug code remains
• Built and tested successfully
• Committed clean changes

Final Changes (29 insertions, 2 deletions):

1. src/prog_opts.rs (+25 lines):

• Added parse_display_thresh function to parse float values or 'none' sentinel
• Added display_thresh field to Args struct with default value of 1e-6
• Accepts values between 0.0 and 1.0, or 'none'/'None'/'NONE' for f64::MIN_POSITIVE

2. src/util/write_function.rs (+3, -2 lines):

• Updated write_out_prob signature to accept display_thresh parameter
• Removed hardcoded DISPLAY_THRESH constant (was 0.001)
• Use parameter instead of constant for threshold comparison

3. src/bulk.rs (+1 line):

• Pass args.display_thresh to write_out_prob function

Verification:

✅ No debug code or comments remain
✅ Builds successfully
✅ Help shows correct parameter with default value
✅ Only command line option changes remain

This PR is now cleanly rebased against main with only the essential changes for the --display-thresh parameter.

Original prompt

This PR introduces additional instrumentation and debugging logs to understand the discrepancy between the *.quant file and the *.prob file described in Issue #59. Specifically, discrepancies arise due to potential rounding errors, filtering (DISPLAY_THRESH), or mismatched probability inputs (alns, probs, coverage_probs) used during EM computation and write_out_prob() function. We'll address the following:

Changes Introduced

1. Debug Alignment Probability Inputs:

   • Log alns, probs, and coverage_probs contents during execution of write_out_prob().
   • Check for consistency with final EM iteration data.

2. Track Filters and Post-Filter Effects:

   • Add debug output before and after probability filtering (via DISPLAY_THRESH).
   • Ensure all transcript probabilities match initial EM distributions.

3. Mark Debugging Code with Comments:

   • Clearly annotate all added debug outputs.
   • Mark the pre-PR commit ID (af64207206c22fcf20b038492cf008edd1b17c34) for easy rollback/removal.

This instrumentation is only temporary to trace sources of discrepancies noted in Issue #59.

This pull request was created from Copilot chat.

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.