The Astra Suite

One Sequence In. The Whole Protein Out.

Astra is Orbion's family of protein AI models. Each reads a sequence and returns a different layer of understanding — structure confidence, function, modifications, binding, topology, and stability. Together, they power the platform.

StructureFunctionModificationsBindingTopologyStability

The Models

Every Read on a Protein

Ten models, each trained for one job and evaluated on public baselines or prospective experiments. Several are documented in open preprints; explore any of them below.

Preprint

AstraPTM2

Modifications

Flags where 39 types of post-translational modification are likely — phosphorylation, glycosylation, ubiquitination and more — at single-residue resolution.

39 PTM TypesPer-Residue

Preprint

AstraROLE2 & AstraSUIT2

Function & Suitability

A ten-factor profile: what a protein does (EC, GO, pathways) and how to work with it (cofactors, domains, host, membrane type, localization) — even for de-novo proteins.

FunctionSuitability

Preprint

AstraBIND

Binding

Locates ligand-binding residues and retrieves likely ligands from a 2M+ panel across 17 categories — combining sequence, structure, and homology.

Binding ResiduesLigand Types

AstraUNFOLD

Topology & Disorder

Membrane topology, intrinsic disorder, and amyloid-aggregation propensity — read residue by residue, from sequence alone.

TopologyDisorderAggregation

AstraDTM & AstraDDG

Stability

How a single mutation shifts melting temperature (ΔTm) and folding free energy (ΔΔG) — so you can rank the changes most likely to stabilize a target.

ΔTm (°C)ΔΔG (kcal/mol)

AstraYIELD

Expression Yield

Ranks a construct panel by how well each will express — validated prospectively and blind — so bench time goes to the designs most likely to produce material.

Yield RankConstruct Triage

AstraINTERACTION

Interactions

Scores whether two partners interact — peptide–MHC, receptor–G-protein, kinase–substrate, protein pairs — from sequence, as a confidence-gated pre-filter.

pMHCCouplingPPI

AstraAFFINITY

Binding Affinity

Predicts protein–small-molecule affinity to rank a compound series for a known target — coarse triage that holds up on a time-split, not a precise Kd.

Affinity RankSequence + Ligand

…and more in the platform

Structure confidence (pLDDT), developability, and functional-state models power the rest of Orbion.

Explore the Platform

One System

Built to Work Together

The models aren't siloed endpoints. Inside the Orbion platform they run in sequence — characterize a protein top to bottom, then score the mutations and constructs worth making — so a single upload turns into a decision.

See the Platform

Pipeline

Characterize

Structure, PTMs, Function, Binding, Topology

Stabilize

Rank Mutations by ΔTm / ΔΔG

Design

Score Constructs for Expression Yield

Decide

The Handful Worth Taking to the Bench

Put Astra on Your Targets

Every Astra model runs inside the Orbion platform. Bring a sequence and get the full read-out — structure, function, modifications, binding, and stability.

Book a Demo Explore the Platform