Decision Engine

The construct
builds itself.

From disease name to optimized AAV vector in 6 deterministic steps. No black boxes. Every decision traceable to a primary reference.

Design Pipeline

01

Disease Input

Natural language disease query → normalized ID lookup with fuzzy matching

02

Gene Resolution

Primary gene identification, modifier genes, associated therapeutic targets

03

Target Ranking

Score genes by role (primary/modifier), AAV compatibility, and clinical significance

04

Tissue Mapping

Disease → target tissues → optimal promoter + serotype + route selection

05

Construct Assembly

Component-level sizing: ITRs + promoter + gene + enhancer + polyA + immunomod

06

Simulation

PK/PD modeling → expression kinetics → immune response → therapeutic window

Disease Explorer

Select a disease to see its gene targets and affected tissues. Live from chaos.db.

Gene Database

0 genes with exact coding sizes from primary references.

GeneFull NameLocusCoding (bp)Mini-GeneSignificance

Tissue Decision Matrix

Priority-ranked selections from 0 promoter and 0 serotype mappings.

TissuePromoterSizeSerotype

Construct Templates

0 pre-computed reference designs.

Tool Modules

Three Python modules. Ten tool functions. Claude is the reasoning engine — tools provide data.

analysis.py
lookup_gene()

Gene info with coding size, chromosomal location, mini-gene availability

analyze_disease()

Disease profile with primary gene, inheritance, target tissues, related genes

rank_therapeutic_targets()

Score and rank all genes for a disease by therapeutic potential

design.py
check_aav_compatibility()

Gene vs 4,700bp AAV limit with mini-gene and dual-AAV strategies

select_promoter_for_tissue()

Priority-ranked tissue → promoter from 23-row decision table

design_construct()

Full construct assembly with auto-sizing, fallbacks, capacity validation

simulator.py
simulate_therapy()

Day-by-day PK/PD + immune + NAb simulation for any combination

simulate_from_db()

Run simulation from a named construct template

compare_serotypes()

Head-to-head comparison of all 6 serotypes

Data Layer

0 genes, 0 diseases, 0 regulatory elements, 0 templates loaded.

0
Genes
0
Diseases
0
Elements
0
Promoters
0
Promoter Maps
0
Serotype Maps
0
Templates
0
Relationships