SDD Wiki


title: “PHGDH Project Brief” updated: 2026-04-18 source_of_truth: /Users/jakeclaw/.openclaw/workspace/project_info.md (Master-Blueprint)

PHGDH Project — what we know (distilled)

The target

PHGDH = D-3-phosphoglycerate dehydrogenase

  • UniProt: O43175 (human)
  • ChEMBL target: CHEMBL2311243 (from scavenger)
  • Function: rate-limiting enzyme in the serine biosynthesis pathway (3-PG → 3-phosphohydroxypyruvate)
  • Cofactor: NAD+ / NADH
  • Quaternary structure: tetramer
  • Key domains: substrate-binding (SBD), nucleotide-binding (NBD), allosteric-serine (ACT) regulatory domain, regulatory-binding domain (RBD)

Why this target (project thesis)

  1. Neurodegeneration angle (primary). PHGDH is dysregulated in Alzheimer’s — evidence from proteomic and transcriptomic studies already indexed in our wiki (phgdh-dysregulation-in-alzheimers-disease-pathogenesis, phgdh-and-amyloid-beta-interactions).
  2. Allosteric / RBD sites, not the active site. Active-site orthosteric inhibitors (NCT-503, CBR-5884) exist but have selectivity problems. Our thesis is that allosteric and RBD modulators can offer finer regulatory control — up OR down — for CNS application.
  3. Serine → D-serine → NMDA modulation. Serine biosynthesis is linked to D-serine (NMDA co-agonist). Tuning PHGDH tunes synaptic signaling, relevant for Alzheimer’s and other neurodegenerative diseases.

Structures we care about

PDBFormNotes
4NJNapo human PHGDH tetramerreference apo
6RIHwith orthosteric ligandactive-site benchmark
7S3Rwith allosteric pocket boundallosteric reference (if matches our thesis)

Others to download as G-005 progresses.

Tools already in place

  • Data scavenger — daily ChEMBL pull to ~/workers/data/phgdh/YYYY-MM-DD.jsonl (2613 rows of bioactivity in today’s file — SMILES, pChEMBL, standard_type, assay IDs)
  • Wiki (Hermes-built) — 17 PHGDH pages already in ~/wiki/general/ and ~/.openclaw/workspace/topics/ covering: structure, SBDD, virtual screening, drug-discovery literature, deep-learning for binding affinity, serine pathway linkages, AD therapeutic strategies
  • Cheaha-HPC skill — templates for CPU, single-GPU, multi-GPU, and array sbatch; wrappers for submit / status / cancel
  • UAB VPN — openconnect + credentials in place; W0 can VPN on demand to reach internal UAB resources

Near-term milestones (from BACKLOG.md)

  • G-002: pChEMBL distribution figure (local — matplotlib)
  • G-003: top-20 inhibitors table (local — pandas)
  • G-004: PubMed sweep for allosteric/RBD literature (Hermes wiki-builder)
  • G-005: PDB structure download + binding-site annotation (local — BioPython)
  • G-006/G-007: Cheaha workspace setup + virtual-screening sbatch
  • G-008: wiki page phgdh-scavenger summarizing the pipeline
  • G-009: paper intro draft (Chrome-Claude)
  • G-010: repo README + pipeline diagram
  • G-011: January 2025 AI-drug-discovery digest

Success criteria (project level)

(Copied from PROJECT_PLAN.md for easy reference.)

  • ≥ 30 days of PHGDH bioactivity data accumulated
  • ≥ 100 source-validated PHGDH wiki pages
  • ≥ 1 SBDD virtual-screening run completed on Cheaha
  • Paper draft: intro + methods + preliminary results sections
  • Slide deck: 20+ slides
  • ≥ 30 consecutive autonomous days (only Duo pushes from Jake)
  • 0 unresolved incidents in RESOLUTION_LOG for 7+ days

Note on project name drift

The project_info.md master-blueprint frontmatter says “PHGDH-Allosteric-RBD-Binder” but the body says “PHAGE”. That was a Gemma 4 tokenizer glitch when the model drafted the body. PHGDH is correct; PHAGE is noise. This brief is canonical.