Skip to content

Latest commit

 

History

History
93 lines (51 loc) · 4.73 KB

File metadata and controls

93 lines (51 loc) · 4.73 KB

Permea Manifesto

Opening Thesis

Delivery is one of the bottlenecks for next-generation therapeutics.

New therapeutic modalities can generate candidates faster than they can reliably deliver them to the right biological context. Peptides, RNA systems, mRNA designs, protein therapeutics, gene-editing payloads, and targeted biologics all depend on delivery. Yet delivery science is still fragmented across datasets, papers, predictors, assays, and internal pipelines.

Permea exists to make delivery evidence benchmarkable, reproducible, and reusable.

Permea is building the open execution layer for delivery engineering: a shared technical foundation for delivery datasets, benchmark tasks, evidence cards, run manifests, candidate reports, and reproducible dry-lab workflows.

Why Delivery Needs a New Layer

The field already has valuable predictors, datasets, reviews, and experimental programs. They matter, but they are not enough by themselves.

Delivery tasks need shared benchmark definitions. A predictor result is difficult to interpret if the task, label source, split policy, metric set, provenance, and limitations are unclear.

Candidate selection needs evidence-backed dry-lab workflows before wet-lab. Experimental work remains essential, but the path into experimental work should be more transparent than a spreadsheet, a paper trail, or an opaque model score.

Literature evidence should become structured and reusable. Claims, assays, sequences, labels, limitations, and source context should be converted into reviewable evidence objects.

Negative and failed evidence should eventually accumulate in structured form. Delivery engineering will improve faster if the field can learn not only from successful candidates, but also from candidates that failed under clear conditions.

What Permea Is Building

Permea is organized around five public-facing building blocks.

Delivery Dataset Commons

A contribution system for delivery-relevant dataset cards, evidence cards, source attribution, readiness levels, and review status. The commons begins with BBB-oriented peptide evidence and expands toward CPP and membrane penetration, localization and targeting, RNA and mRNA delivery-adjacent tasks, and literature-derived evidence graphs.

Benchmark Execution Layer

A reproducible execution layer for defining benchmark tasks, assembling datasets, running baselines, evaluating metrics, tracking provenance, and exporting outputs. The goal is to make delivery tasks executable and comparable.

Dry-Lab Candidate Prioritization

A workflow for sequence-first candidate prioritization before experimental follow-up. Users should be able to provide candidate sequences, choose a delivery task, run a benchmark, review rankings and evidence cards, inspect risk flags, and export a reproducible package.

Research Intelligence Layer

A structured way to turn papers, claims, datasets, and benchmark summaries into reviewable evidence objects. The goal is not to replace scientific review; it is to make evidence extraction, claim checking, and candidate explanation more systematic.

Open Contribution Model

Permea should be built with researchers, developers, computational biologists, delivery scientists, and biotech teams. The field needs shared contribution objects, not only closed one-off analyses.

What Permea Is Not

Permea is not claiming to have solved delivery.

Permea is not claiming wet-lab validation.

Permea is not claiming clinical performance.

Permea is not claiming universal delivery prediction.

Permea is not claiming state-of-the-art status.

Permea is not claiming maturity comparable to AlphaFold.

Permea is a benchmark-first, evidence-first infrastructure program. Its current public claim level is computational and bounded.

AlphaFold-for-Delivery Language

Permea is building the open execution layer required for delivery engineering to become an AlphaFold-scale computational field.

AlphaFold-for-Delivery is an ambition and infrastructure direction, not achieved status.

The analogy is about field transformation: making a hard biological problem more benchmarkable, reproducible, explainable, and executable before experimental follow-up. It is not a claim that delivery has already been solved.

Community Invitation

Permea invites researchers, developers, computational biologists, delivery scientists, and biotech teams to help build an open delivery evidence commons.

Useful contributions include:

  • dataset cards
  • benchmark tasks
  • evidence cards
  • feature descriptors
  • baseline models
  • run manifests
  • reproducible workflows
  • candidate report formats
  • review workflows

The long-term goal is a community where delivery evidence becomes easier to inspect, compare, reproduce, and extend.

Closing

Make delivery benchmarkable.