PTM-EDA: PhosphoSitePlus LTP-tier stratification to test functional-site vs detection-bias signal

## Status (2026-05-31)

Backlog research/EDA task — **not started; blocked on PSP data access** (PhosphoSitePlus requires registration/license). The PTM module now exists in code (`hvantk/algorithms/ptm/{annotate,constraint,analysis}.py`, CLI `hvantk/tools/ptm/ptm_cli.py`); `clinvar_ts_ptm.ht` / `is_ptm_site` are produced there. Per the PTM EDA, the **pathogenicity track closed negative**; the **population-constraint AF-depletion signal is the live finding** this issue aims to disambiguate. Begin only once a PSP-access decision is made.

## Context

Current PTM annotations in `clinvar_ts_ptm.ht` come from **PeptideAtlas** and **CPTAC** — both proteomics-sourced catalogs with no functional-evidence tier. Every detection, whether LTP-validated or one-shot mass-spec observation, gets the same `is_ptm_site` flag.

The proteome-wide ~20–30× AF-depletion signal for PTM variants could be:

- **functional-site biology** — PTM residues under stronger negative selection because they modulate protein function;
- **detection-bias artefact** — long, abundant proteins get sampled more by mass-spec, so the "PTM" label enriches for already-constrained proteins.

Without a functional-evidence tier, we can't distinguish these.

## Goal

Use **PhosphoSitePlus (PSP)** `LT_LIT` (low-throughput literature count) as a functional-evidence tier and re-run the PTM vs non-PTM AF-depletion analysis stratified by that tier.