Feature Request: Support PTM localization for non-phosphorylation modifications

# Feature Request: Support PTM localization for non-phosphorylation modifications

## Overview

onsite is almost entirely **hardcoded for phosphorylation (modification mass 79.96633 Da)** localization. To support other modifications (e.g., acetylation, methylation, oxidation, ubiquitination), a systematic improvment of the code is required.

Below is a layer-by-layer analysis of the phosphorylation-specific hardcoding in the current codebase.

---

## Global Issues (ordered by severity)

### 1. Modification name "Phospho" hardcoded everywhere

All three algorithms have the string `"Phospho"` hardcoded as the modification name:

| Location | Code |
|---|---|
| `onsite/phosphors/phosphors.py:19-26` | `MODIFICATION_NAME = "Phospho"`, `MODIFICATION_TYPES = {"S": "Phospho (S)", ...}` |
| `onsite/lucxor/peptide.py:34` | `_TARGET_MOD_NAMES = {"Phospho", "PhosphoDecoy"}` |
| `onsite/lucxor/psm.py:366-393` | String matching for `"(Phospho)"` and `"(PhosphoDecoy)"` |
| `onsite/ascore/ascore.py:84-85` | `sequence_str.count("(Phospho)")` / `sequence_str.count("(PhosphoDecoy)")` |
| `onsite/ascore/cli.py:307` | `"(Phospho)"` counting, `"(PhosphoDecoy)"` counting |
| `onsite/phosphors/cli.py:333` | `"(Phospho)"` counting |
| `onsite/phosphors/cli.py:403` | `f"{aa}(Phospho)"` and `f"{aa}(PhosphoDecoy)"` pattern matching |
| `onsite/decoy_flr.py:40-41` | `PHOSPHO = "Phospho"`, `PHOSPHO_DECOY = "PhosphoDecoy"` |
| `onsite/onsitec.py:179-181, 485-487` | `target_mods = ("Phospho (S)", "Phospho (T)", "Phospho (Y)")` |
| `onsite/lucxor/cli.py:91` | Default `--target-modifications` is `["Phospho (S)", "Phospho (T)", "Phospho (Y)"]` |

**Impact**: Any modification not named `"Phospho"` (e.g., `"Acetyl"`, `"Methyl"`, `"Oxidation"`) will not be recognized as a target for localization.

### 2. Modifiable residues hardcoded to S/T/Y

All three algorithms only recognize Ser/Thr/Tyr as possible modification sites:

| Location | Code |
|---|---|
| `onsite/ascore/ascore.py:62-64` | `isPhosphoSite(): return residue in ["S", "T", "Y"]` |
| `onsite/phosphors/phosphors.py:26` | `POTENTIAL_SITES = set(["S", "T", "Y"])` |
| `onsite/lucxor/psm.py:243` | `self.target_mod_map = {"S": True, "T": True, "Y": True}` |
| `onsite/lucxor/peptide.py:601-603` | `if aa in ["S", "T", "Y"]` |

**Impact**: Residues commonly modified by other PTMs — K (acetylation, ubiquitination), M (oxidation), C (alkylation), etc. — will never be considered as candidate sites for localization scoring.

### 3. Modification mass hardcoded to 79.96633 Da (HPO₃)

| Location | Code |
|---|---|
| `onsite/lucxor/constants.py:51` | `PHOSPHO_MOD_MASS = mass_provider.get_phospho_mass()` ≈ 79.96633 |
| `onsite/lucxor/constants.py:58-66` | `AA_MASSES.update({"s": AA_MASSES["S"] + PHOSPHO_MOD_MASS, ...})` |
| `onsite/lucxor/constants.py:97` | `DECOY_MASS = PHOSPHO_MOD_MASS` |
| `onsite/phosphors/phosphors.py:740` | `_PHOSPHO_MOD_MASS = 79.96633` |
| `onsite/ascore/ascore.py:697-706` | Sets `"Phospho"` modification via PyOpenMS (implicitly using phospho mass) |
| `onsite/phosphors/phosphors.py:1188-1192` | Mass-based filtering for phospho modification mass |

**Impact**: Correct masses for other modifications (e.g., acetylation +42.010565 Da, oxidation +15.994915 Da, methylation +14.015650 Da) cannot be used for theoretical spectrum calculation or fragment matching.

### 4. Neutral losses hardcoded to H₃PO₄ (-97.97690 Da)

The LucXor CID fragmentation model relies heavily on the characteristic neutral loss of phosphorylation:

| Location | Code |
|---|---|
| `onsite/lucxor/constants.py:111-112` | Default `neutral_losses = ["sty -H3PO4 -97.97690"]` |
| `onsite/lucxor/constants.py:140-149` | `NEUTRAL_LOSSES` dict with hardcoded `"H3PO4"`, `"sty": -97.97690` |
| `onsite/lucxor/peptide.py:548-549` | `if nl.startswith("sty")` for neutral loss assignment |
| `onsite/lucxor/peptide.py:371-386` | Converts lowercase `s/t/y` to PyOpenMS format with phospho neutral loss |
| `onsite/onsitec.py:194, 499` | `neutral_losses=("sty -H3PO4 -97.97690",)` |
| `onsite/lucxor/cli.py:97` | Default `--neutral-losses` is `["sty -H3PO4 -97.97690"]` |
| `onsite/lucxor/constants.py:114` | `decoy_neutral_losses = ["X -H3PO4 -97.97690"]` |

**Impact**: Fragment ion scoring for other modifications uses incorrect neutral loss masses, producing distorted localization scores. Other modifications have different neutral loss signatures (e.g., acetylation loses ketene -42.010565 Da from lysine under CID, but this is not the only or most common neutral loss for acetylated peptides).

### 5. Decoy strategy hardcoded to Alanine (A)

The current false localization rate (FLR) estimation relies on Alanine as the only decoy residue:

| Location | Code |
|---|---|
| `onsite/decoy_flr.py:42-43` | `TARGET_RESIDUES = set("STY")`, `DECOY_RESIDUE = "A"` |
| `onsite/ascore/ascore.py:57-59` | `isPhosphoDecoySite(): return residue == "A"` |
| `onsite/phosphors/phosphors.py:23-24` | `"A": "PhosphoDecoy (A)"` |
| `onsite/lucxor/constants.py:168-172` | `DECOY_AMINO_ACIDS = {"S": "A", "T": "V", "Y": "F"}` |

**Impact**: The decoy-amino-acid method couples Alanine to the assumption that S/T/Y are the only target residues. For other modifications, Alanine may not be a suitable decoy; a different decoy residue or an entirely different FLR strategy may be needed.

### 6. Lowercase letter convention hardcoded to s/t/y

LucXor uses lowercase letters to encode modified residues internally, but only defines s/t/y (and a/m):

| Location | Code |
|---|---|
| `onsite/lucxor/peptide.py:371-386` | `_to_pyopenms_format()` only recognizes `s→S(Phospho)`, `t→T(Phospho)`, `y→Y(Phospho)`, `a→A(PhosphoDecoy)` |
| `onsite/lucxor/psm.py:1075-1083` | `_get_mod_map()` checks `aa.upper() in ["S", "T", "Y"]` and `"a"` for Alanine decoy |
| `onsite/lucxor/psm.py:1199-1208` | `_get_mod_positions_from_perm()` checks `aa.upper() in "STY"` and `"a"` |
| `onsite/lucxor/psm.py:1753-1769` | `convert_sequence_to_standard_format()` only checks for lowercase `s/t/y` and `"a"` |
| `onsite/lucxor/constants.py:58-66` | `AA_MASSES` only adds mass entries for `s/t/y/a/m` |

**Impact**: No lowercase letter mappings exist for other modification sites (e.g., `k` for acetylated lysine, `m` for oxidized methionine — though `m` is defined for oxidation, it's not integrated into the full pipeline), so the internal sequence representation cannot propagate other modification information.

Start by tracing the phosphorylation-specific constants and defaults in onsite/lucxor/constants.py, onsite/lucxor/cli.py, and onsite/onsitec.py, then compare the site and mass handling in the listed ascore, phosphors, lucxor, and decoy_flr modules. Map how modification names, residues, masses, neutral losses, and decoys flow through all three algorithms. Done means non-phosphorylation PTMs can be configured and localized consistently without phosphorylation-only assumptions.

Written by the indexing model from the issue text.