| Inspect the MACS3 source tree and public reproduction. Repair the scientific |
| peak-refinement workflow so that |
| valid ATAC-seq candidate regions can be processed together with a sparse |
| fold-change signal track across legitimate breakpoint and resolution patterns. |
|
|
| Preserve the scientific meaning of genomic interval boundaries, quantitative |
| scores, and summit locations for ordinary cases. The scored workflow in this |
| task follows the historical array-backed `bedGraphTrackI` path used by the |
| upstream regression. The repair must remain valid across nested, adjacent, and |
| breakpoint-misaligned intervals. Do not hard-code the public chromosome names, |
| coordinates, values, case names, or output counts. |
|
|
| Within each chromosome, score segments are ordered scientific evidence. Once a |
| segment's evidence has been consumed by and attributed to a completed refined |
| peak, it must not be counted or attributed again to a later overlapping or |
| nested candidate. A candidate with no remaining supporting evidence need not |
| emit a record. This is a result-level attribution requirement; the traversal, |
| state representation, and implementation strategy are intentionally left open. |
|
|
| Run `python reproduce.py` from this task directory to build the local MACS3 |
| runtime and inspect the public symptom. You may edit the source tree and run |
| the reproduction repeatedly. Do not add generated Cython build products to |
| your patch. |
|
|
