pycan-ultra 0.1.0

Comprehensive agentic AI platform for end-to-end cancer research and precision oncology

pycan_ultra

pycan_ultra is a deep, modular, and agentic Python platform for end-to-end cancer research, translational analytics, and precision oncology decision support.

Built as a full-stack research engine, it unifies autonomous multi-agent reasoning, multi-omics integration, survival modeling, drug prioritization, evidence retrieval, and publication-grade visualization in one package.

Why pycan_ultra

Most oncology stacks require stitching many isolated tools. pycan_ultra provides a coherent, typed, and reproducible architecture that supports both exploratory science and production pipelines.

Key strengths

• Agentic intelligence layer: orchestrates genomics, transcriptomics, survival, pathology, drug-discovery, reporting, and memory agents from a single objective.
• Multi-omics depth: supports TCGA/GEO/cBioPortal ingestion patterns and harmonized modality processing for genomics, transcriptomics, spatial, methylation, proteomics, and metabolomics.
• Precision oncology engine: biomarker ranking, mutation signatures, neoantigen ranking, immunotherapy scoring, resistance forecasting, and patient-ready summary generation.
• Modeling stack: survival, multimodal fusion, graph ranking, anomaly detection, ctDNA risk, explainability, and model registry primitives.
• Knowledge and evidence system: OncoKB/CIViC-style abstractions, variant annotation, evidence ranking, and literature-aware answer generation.
• Pipeline-first design: discovery, patient, cohort, longitudinal, NGS, spatial, IO, and benchmark pipelines with scheduler-ready interfaces.
• Clinical-grade foundations: structured schemas, explicit exceptions, caching, logging, deterministic utilities, and test scaffolding.
• Offline-capable operation: rule-based fallback paths for environments without LLM/API access.

Feature map

Layer	What it provides
Agents	Autonomous planning, tool-call envelopes, reflection loops, state persistence
Omics	Typed loaders, harmonization helpers, integration and validation contracts
Precision	Clinical scoring modules and recommendation-ready outputs
Models	Research model primitives and deploy/export-oriented interfaces
Pipelines	One-function workflows for discovery, patient, and cohort operations
Knowledge	Evidence retrieval, annotation, and ranking for grounded decisions
Viz	OncoPrint, KM points, landscape, pathway, spatial, and reporting visuals
Core/Utils	Config, logging, cache, schemas, auth/rate-limits, async/parallel helpers

Installation

pip install pycan-ultra

Enable optional stacks:

pip install "pycan-ultra[omics,dl,agents,viz,dev,docs]"

Quick start

from pycan_ultra import __version__
from pycan_ultra.pipelines.discovery_pipeline import run_discovery

print("pycan_ultra", __version__)
result = run_discovery("BRCA", 1200)
print(result["targets"]) 

CLI

pycan --help
pycan discover --cancer BRCA --omics tcga

Packaging and distribution

This project uses Hatchling via pyproject.toml.

Build artifacts:

python -m build

Upload (when token is configured):

python -m twine upload dist/*

License

Apache-2.0