RIA-01 WP4 — Safety Risk Matrix

⚠ WP4 Headline

Two safety lines define the retina-first program risk envelope:

Tazemetostat secondary-malignancy signal — FDA boxed warning in pediatric epithelioid sarcoma. Manageable via cyclic / low-dose architecture and adult-only enrollment, but cannot be ignored.
Cyclic OSK / AAV re-dosing tension — cyclic OSK is the safety architecture from Belmonte / Browder. AAV is biologically one-shot. This is a real architectural conflict, not a safety failure, and Atlas Bio's engineered-capsid program is the natural mitigation.

Beyond those two, the program is unusually safety-clean for a gene-therapy + small-molecule combination: vitamin C / α-KG priming, 4D-R100 intravitreal delivery, and pulsed Tazemetostat each have prior human exposure or compartment-isolated routes that bound the risk envelope.

1. Risk categories used in this matrix

Category	What it captures
DD — Dedifferentiation	Risk of driving cells past partial reprogramming into full pluripotency (loss of cell identity)
TR — Teratoma	Tumor formation from fully reprogrammed cells
ON — Oncogenic transformation	Pathway activation that drives malignant transformation independent of pluripotency
OT — Off-target epigenetic	Modulation of chromatin marks outside the intended axis
IM — Immune / inflammatory	Innate or adaptive immune response, including anti-AAV immunity
CH — Chronic dosing concerns	Cumulative toxicity, dependency, or pathway exhaustion with repeated exposure
CA — Co-administration	Mechanistic interaction with common comorbid drugs or supplements
AE — Absence of evidence	Material safety question on which the published literature is silent (a risk in itself)

LOW

MED

HIGH

UNK (absence of evidence)

N/A

2. Chemistry candidates — per-candidate risk grid

Candidate	Role	DD	TR	ON	OT	IM	CH	CA	AE	Net
Vitamin C / ascorbate	Prime	LOW	LOW	LOW	LOW	LOW	LOW	LOW	LOW	LOW
α-Ketoglutarate	Prime	LOW	LOW	LOW	MED	LOW	LOW	MED	MED	LOW
Tazemetostat	Maintain	LOW	LOW	HIGH	LOW	LOW	MED	MED	LOW	MED
RepSox / E-616452	Prime (alt)	LOW	LOW	LOW	MED	LOW	MED	MED	MED	LOW
CHIR99021	Prime (cocktail)	LOW	LOW	MED	LOW	LOW	MED	MED	LOW	MED
Tranylcypromine	Prime (cocktail)	LOW	LOW	LOW	MED	LOW	MED	HIGH	LOW	MED
VPA	Prime (cocktail)	LOW	LOW	LOW	HIGH	LOW	HIGH	MED	LOW	HIGH
DZNep	Maintain (legacy)	LOW	LOW	MED	HIGH	LOW	HIGH	MED	MED	HIGH
OICR-9429 (WDR5)	Future / probe	LOW	LOW	LOW	MED	LOW	UNK	UNK	UNK	MED
EPZ-5676 (DOT1L)	Future / maintenance	LOW	LOW	MED	LOW	LOW	MED	LOW	MED	MED
KAT8 activator (TBD)	Future / maintenance	UNK	UNK	UNK	UNK	UNK	UNK	UNK	HIGH	UNK
5-Aza-2'-deoxycytidine	Excluded	LOW	LOW	HIGH	HIGH	LOW	HIGH	MED	LOW	HIGH

"Net" = RIA-01's qualitative composite, weighted to favor categories most consequential to the proposed retina-first architecture. Chronic risk (CH) and absence-of-evidence (AE) are weighted higher than for typical drug-development reviews because partial reprogramming is intrinsically cyclic / chronic.

3. Delivery vehicles — per-capsid risk grid

Vehicle	Indication	Pre-existing immunity	Re-dosability	Off-target tissue exposure	Vector-related toxicity	Net
4D-R100 intravitreal	Retina / RGC	LOW (compartment-isolated)	MED (eye allows some re-dosing)	LOW	LOW	LOW
PHP.eB IV systemic	CNS (mouse-optimized)	UNK in humans	HIGH (NAb formation)	MED	MED	HIGH (mouse-only validation)
AAV9 IV systemic	CNS, heart, multi-tissue	HIGH (30–60% adult NAb prevalence)	HIGH	HIGH (broad tropism cuts both ways)	MED	HIGH for systemic; LOW for intrathecal
4D-C102 IV cardiac	Cardiomyocyte	MED	HIGH	LOW (cardiac-targeted)	MED	MED
LNP-mRNA	Multi-tissue (liver-dominant systemic)	LOW (no NAb durability)	LOW (re-dosable)	HIGH (liver-dominant biodistribution)	MED (innate immune activation)	MED

4. OSK cassette content — risk by transcription factor included

Cassette	Reprog efficacy	Oncogenic risk	Dedifferentiation risk	RIA-01 recommendation
OSKM (O+S+K+M)	Highest	HIGH (c-MYC)	HIGH	Reject for partial reprogramming — oncogenic ceiling unacceptable
OSK (no MYC)	High	MED	MED	RECOMMENDED for first-in-human partial reprogramming
OS (no K, no M)	Lower but viable when paired with chemical priming	LOW	LOW	Candidate for chemical-prime-heavy protocols; lower expected efficacy
OSKMLN (Sebastiano)	High, transient	MED (c-MYC briefly present)	LOW (mRNA transient)	Native mRNA protocol; choose only if Turn-Bio-style mRNA delivery is the platform

SAFETY POSITION

Drop MYC from the cassette for any AAV-based partial reprogramming program. The Ocampo 2016 / Browder 2022 / Lu 2020 lineage all use cyclic OSK or doxycycline-inducible OSKM with strict on/off control. The oncogenic ceiling of c-MYC is the single largest dial-able safety lever in the architecture.

5. Architecture-level risk per WP3 pairing

Pairing	Compartment risk	Cyclic-dosing risk	Re-dose risk	Co-admin risk	Aggregate architecture risk
#1 Retina-first: Vit C + α-KG → 4D-R100 AAV-OSK → Tazemetostat	LOW (intravitreal)	LOW (oral chemistry; pulsed AAV)	MED (eye re-injection possible)	MED (NAD+ supplements; tyramine for any Parnate variant)	LOW
#2 CNS: Vit C + α-KG → PHP.eB / AAV9 AAV-OSK → Tazemetostat	HIGH (systemic AAV)	LOW (oral chemistry)	HIGH (AAV NAb)	MED	HIGH
#3 KAT8 maintenance (future)	LOW	UNK (KAT8 activator unknown)	MED	HIGH (NAD+ antagonism explicit)	MED (conditional on tool)

6. Co-administration risk register

Concomitant	Mechanism of interaction	Direction	Risk	Mitigation
NMN / NR (NAD+ precursors)	Raise NAD+ → activate SIRT6 → deacetylate H4K16	Antagonizes KAT8 maintenance layer (WP1.3 conflict)	HIGH	Counsel patients off NAD+ supplements during maintenance phase
High-dose nicotinamide (NAM)	Sirtuin product inhibition	Could augment H4K16ac (opposite direction from NMN/NR)	MED	Capture in baseline supplement history; not actively recommended
Vitamin C megadose IV	H₂O₂ via Fenton chemistry at high local concentrations	Iron-rich tissues at risk; eye / RPE has iron stores	MED	Oral priming doses (1–3 g/day) only; not IV megadose
MAOIs / SSRIs (any Parnate variant)	MAO-A/B inhibition + serotonergic agents = serotonin syndrome	Direct adverse interaction	HIGH	Switch tranylcypromine to selective LSD1-i (ORY-1001) if cocktail requires LSD1 axis
Anti-AAV9 neutralizing antibodies	Pre-existing or induced humoral immunity	Reduces transduction; can cause inflammation	HIGH (systemic) / LOW (intravitreal)	NAb titer screen pre-dose; immunosuppressive pre-medication if appropriate
Topical or systemic steroids	Modulate inflammatory milieu	May suppress AAV-induced inflammation (helpful) and partial-reprogramming SASP signal	MED	Capture in concomitant medication record; do not exclude unless clinically inappropriate
Cancer history / current malignancy	Tazemetostat secondary-malignancy signal; OSK pulse oncogenic ceiling	Exclusionary, not merely interactive	HIGH	Exclusion criterion for any human trial

7. Absence-of-evidence flags (risks because the literature is silent)

AE-1 · KAT8 activator chronic safety

No clean KAT8 activator exists; therefore no chronic-safety data exists. The KAT8 program (WP2.1) must build its own safety dataset. This is the largest AE flag in the entire study.

AE-2 · Cyclic OSK in primates

The Belmonte / Browder / Lu evidence base is mouse-dominant. Cyclic OSK in non-human primates has limited disclosure. The first human trial will be the first cyclic primate exposure at therapeutic doses.

AE-3 · Tazemetostat in non-oncology indications

Tazemetostat's safety database is cancer-patient-derived. Aging-indication patients (healthier baseline, longer expected exposure) may have a different secondary-malignancy signal magnitude.

AE-4 · 4D-R100 in long-term re-dosing

4DMT's R100 capsid clinical experience is at the first-dose level. Multi-dose intravitreal partial-reprogramming protocols (every 6–12 months hypothesized) are not validated.

AE-5 · OSK partial reprogramming and ocular tumor formation

Lu 2020 reports no teratoma in the time-frames studied. Lifetime tumor surveillance data does not exist for any partial-reprogramming protocol.

AE-6 · α-KG long-term ophthalmic effects

α-KG supplementation literature focuses on systemic / metabolic endpoints. Ophthalmic-specific long-term data are sparse.

8. Net safety verdict on the recommended retina-first program

CLEARED LANES

Vitamin C / α-KG priming, intravitreal 4D-R100, OSK cassette (drop MYC), and pulsed Tazemetostat each pass independently. The compartmentalized eye route is the single most important risk-reducing architectural choice.

MATERIAL SAFETY LINES (managed, not eliminated)

Tazemetostat secondary-malignancy boxed warning — manage via cyclic / low-dose; adult enrollment; cancer-history exclusion. AAV re-dosability — manage via engineered capsid pipeline and surface-modified variants; reserve PHP.eB / AAV9 systemic for later expansion.

REQUIRED PROSPECTIVE DATA

Cyclic OSK primate safety; 4D-R100 multi-dose intravitreal safety; aging-indication Tazemetostat exposure. These are appropriate IND-enabling safety questions and not gating issues for the computational study.

RIA-01 net verdict: the retina-first program (WP3 Pairing #1) has an unusually favorable computational safety profile for a gene-therapy + small-molecule combination. The two material lines (Tazemetostat boxed warning, AAV re-dosability) are managed by architecture and patient selection, not by mechanism change. Forward planning for WP5 endpoints and WP-final synthesis can proceed.

Safety risk matrix for the retina-first program and its candidate substitutions