Home De Novo Architecture 🧠 NPEAD 🧬 FRSI CIFOG Gates Quantum Wet Lab Control

🧬 CortexQ-FRSI: Functional Regulatory State Inversion

CortexQ is an autonomous scientific reasoning agent. It uses FRSI as its biological reasoning capability — how it thinks about regulatory states, pathway control, and target validity. EGAO, NPEAD, CIFOG, FGAO, and QGCN are its other reasoning tools. MindNetQ governs what CortexQ is allowed to do.

Core Question: If we perturb target X in cancer Y, will pathway Z move away from the malignant state?

FRSI does not rank genes by expression. It scores target-cancer pairs by predicted ability to invert malignant regulatory states, then emits falsifiable wet-lab marker predictions.

Pipeline Overview

Cancer Expression Data → ┌─────────────────┐ → Pathway Activity Scores
                         │ Cancer State     │
Normal Tissue Data ────→ │ Encoder          │ → TF Activity Scores
                         └─────────────────┘
                                 │
                                 ▼
                         ┌─────────────────┐
                         │ Discordance     │ → Hidden vulnerabilities
                         │ Detector        │   (high activity, low expression)
                         └─────────────────┘
                                 │
                                 ▼
                         ┌─────────────────┐
DepMap + L1000 ────────→ │ State Inversion │ → Regulatory Inversion Score (RIS)
                         │ Predictor       │
                         └─────────────────┘
                                 │
                                 ▼
                         ┌─────────────────┐
                         │ Marker          │ → 3 qPCR markers + direction
                         │ Predictor       │   (e.g., DUSP6↓, EGR1↓, SPRY2↓)
                         └─────────────────┘
                                 │
                                 ▼
                         ┌─────────────────┐
                         │ Authority Gate  │ → reject | hold | qPCR | protein | pharma
                         └─────────────────┘

Regulatory Inversion Score (RIS)

RIS = Activity Alignment × State Reversal × Marker Coherence
      × Cancer Specificity × Perturbation Evidence
      − Safety Penalty − Contradiction Penalty

Component	What it measures	Source
Activity Alignment	Target connects to active cancer pathways	PROGENy + DoRothEA
State Reversal	Perturbation moves state away from malignant	DepMap + L1000
Marker Coherence	Predicted markers agree with pathway direction	Pathway response genes
Cancer Specificity	Signal stronger in tumor vs normal	TCGA vs GTEx
Perturbation Evidence	Public data supports direction	CMap/L1000
Safety Penalty	Normal-tissue essentiality risk	DepMap + GTEx
Contradiction Penalty	Conflicting evidence, promiscuity	Multi-source

Decision Rules

RIS Score	Action	What happens
< 0.30	Reject	Insufficient evidence for regulatory control
0.30 – 0.50	Hold	Collect more evidence before testing
0.50 – 0.70	qPCR Test	Run 3-marker directional test (~$200)
0.70 – 0.85	Protein Test	qPCR + phospho-protein confirmation (~$800)
> 0.85	Pharma Package	Build evidence bundle for partner review

Predicted Pathway Markers

Each scored target emits 3 qPCR markers with expected direction. These are the cheapest falsifiable test of the FRSI hypothesis.

Pathway	If pathway is ACTIVE in cancer...	After perturbation, expect:
MAPK/ERK	DUSP6↑ EGR1↑ SPRY2↑	DUSP6↓ EGR1↓ SPRY2↓
PI3K/AKT	SGK1↑ TXNIP↓	SGK1↓ TXNIP↑
Wnt/β-catenin	AXIN2↑ MYC↑ CCND1↑	AXIN2↓ MYC↓ CCND1↓
NF-κB	IL6↑ CXCL8↑ NFKBIA↑	IL6↓ CXCL8↓ NFKBIA↓
p53	CDKN1A↓ BAX↓	CDKN1A↑ BAX↑ (reactivation)
Androgen	KLK3↑ FKBP5↑ TMPRSS2↑	KLK3↓ FKBP5↓ TMPRSS2↓
Estrogen	TFF1↑ GREB1↑ PGR↑	TFF1↓ GREB1↓ PGR↓
EMT	VIM↑ ZEB1↑ CDH1↓	VIM↓ ZEB1↓ CDH1↑

Discordance Detection

Key insight: A gene can be modestly expressed but functionally critical. Traditional expression-based target discovery misses these. FRSI's discordance detector finds them.

Type	Meaning	Value
Hidden Vulnerability	Low expression but high functional activity	HIGH — hidden regulatory control point
Concordant	Expression matches activity	Standard target
Overexpressed Bystander	High expression but low functional impact	AVOID — expression artifact

Cancer Coverage

Cancer	Key Pathways	Primary TFs	Status
Prostate	Androgen, PI3K, MAPK, Wnt	AR, MYC, FOXO3, ERG	Active (59 heroes)
Pancreatic	KRAS/MAPK, NF-κB, TGFβ, EMT	MYC, STAT3, SMAD3, NFkB1	Early (7 heroes)
Lung	EGFR, KRAS/MAPK, PI3K, p53	MYC, TP53, E2F1, HIF1A	Scanning
Breast	ER, HER2, PI3K, CDK, EMT	ESR1, MYC, E2F1, STAT3	Scanning

Integration with E-DICE-R Pipeline

physician-rs (scan) → FRSI scoring → CIFOG graph → EGAO authority → generation → docking
                          ↑                                                        │
                          └──────── wet-lab feedback ←────── update_hero ←─────────┘

FRSI writes to each hero:
frsi_score — Regulatory Inversion Score (0-1)
frsi_markers — predicted qPCR markers + expected direction
frsi_action — recommended next action
frsi_discordance — concordant / hidden_vulnerability / bystander

Key Differentiator

Traditional AI Discovery:
• Ranks genes by expression
• Scores by literature count
• Outputs: gene list
• Stops at prediction

CortexQ-FRSI:
• Scores target-cancer-state pairs
• Infers regulatory activity
• Outputs: score + markers + action + modality
• Updates after wet-lab evidence

Live FRSI Results

Loading from API...