A developmental-reference evaluation platform for pre-transplant mDA progenitor cell products.
🌉 Candidate discovery, identity stability, and multidimensional developmental concordance.
Stem-cell-based replacement therapy is an important regenerative strategy for Parkinsonian dopaminergic circuit repair. Pre-transplant mDA progenitor products are evaluated as developmentally staged cells with defined regional identity, fate stability, and subsequent differentiation potential.
BRIDGE uses human embryonic ventral midbrain references to guide candidate-cell discovery, target identity assessment, and multidimensional developmental concordance scoring. The workflow organizes single-cell evidence for quality control, process optimization, and cross-protocol comparison.
| Evaluation layer | Biological focus |
|---|---|
| Developmental reference | Human embryonic ventral midbrain programs as the in vivo baseline. |
| Candidate identity | Calibrated probability, prediction variability, and entropy. |
| Composite Likeness Score | Identity, expression, transferability, neighborhood, trajectory, and regulon concordance. |
| Step | Role | Output |
|---|---|---|
| Step0 | Prepare environment, config, model assets, and run directory. | Ready-to-run workspace |
| Step1 | Map one in vitro .h5ad against a whole-brain reference. |
RG candidate annotations and Step1 report |
| Step2 | Refine mDA progenitor identity with probability and uncertainty. | Candidate-bearing data, thresholds, probability tables, Step2 report |
| Step3 | Quantify developmental concordance with CLS components A-F. | Component scores, weighted CLS, single-dataset and protocol-comparison reports |
pip install git+https://github.com/starvingarc/BRIDGE.git
# or, from a cloned source tree:
pip install -e ".[workflow]"For agent-assisted setup, send this prompt to your coding agent:
Help me install https://github.com/starvingarc/BRIDGE
BRIDGE includes repository-local skills that guide an agent through reproducible Step0-Step3 notebooks. Use the prefix supported by your agent, for example /bridge-step1 or @bridge-step1.
| Step | Skill | Output |
|---|---|---|
| Step0 | bridge-step0 |
Environment, assets, config, and run directory |
| Step1 | bridge-step1 |
Prescreened data, RG candidates, and notebook report |
| Step2 | bridge-step2 |
Identity candidates, thresholds, probabilities, and notebook report |
| Step3 | bridge-step3 |
CLS component scores and protocol comparison |
Full copy-paste demo prompts are in docs/agent_demo.md. Model assets are declared in models/assets.json and fetched separately from public object storage.
from bridge.prescreen import prescreen
from bridge.identity import identify
from bridge.cls import CLSContext, component_A, component_B, component_C, component_D, component_E, component_F, score
from bridge.prescreen.report import write_report as write_prescreen_report
from bridge.identity.report import write_report as write_identity_report
from bridge.cls.report import write_report as write_cls_report, compare_reportsEach step is a Python function that can be used in notebooks or scripts. Report modules provide displayable table/figure helpers and writers for reproducible artifacts under report/.
PYTHONPATH=src pytest -qsrc/bridge/ Python package
configs/ public config templates
models/ model metadata and asset entry point
notebooks/ curated notebook examples; generated notebooks are run artifacts
docs/ workflow documentation and roadmap
.claude/skills/ repository-local Step0-Step3 skills
BRIDGE is research software under active development. If you use it in a study, please cite the repository and include the commit hash used for analysis.