Talk to an Scientist
Talk to an Scientist

Services

Fully automated humanization. Hundreds of thousands of designs, ranked in silico

Lead with the platform’s Al depth. Best for discovery leads evaluating vendors on modernization of the humanization workflow.

  • 127-mAb
    benchmark
  • ~100k
    designs / mAb
  • 7-metric
    liability panel
  • Fully automated
    humanization

Where Discovery Programs Lose Weeks

Why antibody engineering programs stall

Traditional humanization is manual, slow, and blind to downstream liabilities until the wet lab surfaces them.

Three places that pattern quietly costs discovery programs their best candidates.

  1. Manual Humanization

    The bandwidth problem

    • Hand-designed grafts limit to a few candidates
    • Each variant takes weeks to design and rationalize
    • Framework choice is often driven by familiarity
    • One shot; no A/B against alternatives

  2. Liabilities Discovered Late

    The blind-triage problem

    • Dipeptide, patch, and HLA liabilities hide in sequence
    • Wet-lab failures loop back to lead re-selection
    • Weeks of expression and screen lost per bad lead
    • In vivo half-life risk missed until formulation

  3. No Benchmark

    • “Good” liability scores are meaningless in isolation
    • Team debates severity without a reference set
    • Prioritization stalls; sub-optimal leads advance
    • PD program inherits developability risk

Three pillars

How ALTHEA™
rebuilds the workflow

ALTHEA™ is Celltheon’s Al protein-engineering platform. It replaces manual humanization with a fully automated design – in-silico screen – wet lab hand-off pipeline, benchmarked against real commercial mAbs.

  1. Pillar 1

    In-silico liability screen

    • 7-metric developability pipeline per Fv
    • Structure prediction with ensemble ranking
    • In-silico viscosity prediction pre-report
    • Every score is a percentile vs 127 commercial mAbs
  2. Pillar 2

    Fully automated humanization

    • ~100k HC/LC frameworks generated per mAb
    • Ranked by ALTHEA™ + HLA-II 9-mer count
    • Structural liabilities filtered before short-list
    • Top 5 humanized variants returned as report
  3. Pillar 3

    Wet-lab validation loop

    • Transient expression on OmniCHO™
    • SE-HPLC, DSF, HIC, Octet, BVP ELISA panel
    • Half-life, glycan, and mass-variant readouts
    • Client picks final variants from ranked report

In-silico liability panel

The pipeline runs on every candidate before humanization and after. Each metric returns an absolute value and a percentile against a curated 127-mAb commercial
laG benchmark.

  1. Model

    Structure Prediction

    5-seed structural ensemble per Fv. Best model selected by combined confidence score.

  2. Topology

    Numbering & CDR Definition

    IMGT scheme via AntPack. Consistent CDR / framework boundaries across every candidate.

  3. SASA

    Solvent Accessibility

    Per-residue SASA at atom level. Feeds every downstream exposure-weighted metric.

  4. Sequence

    Dipeptide & N-Glyc Liabilities

    14 dipeptide motifs + N-X-[S/T]. Exposure-weighted; CDR hits weighted 2x.

  5. Surface

    Hydrophobic Patches

    Multi-scale NPP with single-linkage clustering. Predicts HIC retention and aggregation risk.

  6. Surface

    Ionic Patches

    Positive / negative surface clusters at physiological pH. Predicts self-association and CEX behavior.

  7. Immuno

    HLA-II Immunogenicity

    9-DRB1-allele scan. CDR-overlap strong binders reported per allele and cumulative strength.

  8. Formulation

    Viscosity Prediction

    In-silico viscosity risk score. Flags candidates likely to trend high at concentration before wet-lab.

Scientists working in the Celltheon laboratory

ALTHEA™
humanization funnel

Every parent mAb is expanded into a design pool of frameworks, then reduced by structural, sequence, and immunogenicity filters. The client sees only the ranked short list — not the compute in between.

  • HC + LC framework designs

    ~100k

    Framework generator across
    germline space

  • ML-modeled structures

    ~3k

    3k HC + 3k LC ranked by
    ALTHEA™ score

  • Filtered by 9-mer HLA-II

    ~300

    Down-selected by non-human
    peptide count

  • Liability-filtered set

    ~100

    50–100 HC and LCs after dipeptide / patch filters

  • Clustered families

    10–20

    10–20 HC and LC diversity-clustered families

  • Top humanized variants

    3–5

    Ranked by ALTHEA™ · returned
    as report

Three Ways to Engage

Engagement models

Liability screening is available as a stand-alone service. Humanization and the full wet-lab validation loop are the primary engagement — most programs come in through the full-loop bundle.

  • Primary offer
    Full Loop — Primary

    Screen → humanize → validate

    In-silico screen + ALTHEA™ humanization + wet-lab validation.

    • One parent in, ranked top 3–5 HC × 3–5 LC out
    • In-silico + wet-lab in one contract
    • OmniCHO™ transient expression included
    • Ranked report + Celltheon recommendation

    Ideal for: programs advancing a lead mAb from mouse to human

    Talk to a Scientist
  • Liability Screen Only

    In-silico developability report

    Score any panel of mAbs against the 127-mAb commercial benchmark, no humanization.

    • Rank any number of candidates
    • Full report + per-clone hotspot maps
    • CDR-only and Full Fv scores returned
    • Feeds triage before wet-lab commit

    Ideal for: teams picking a lead from a discovery panel

    Talk to a Scientist
  • Humanization Only

    ALTHEA™ report without wet-lab

    In-silico humanization + liability panel; client runs wet-lab in-house or via other partner.

    • ~100k → top-candidate down-select
    • Full ranked variant report
    • Sequences ready for gene synthesis
    • Wet-lab handled by client

    Ideal for: teams with strong in-house expression capacity

    Talk to a Scientist
Frequently Asked Questions

Quick answers
to common questions

Quick answers about SUPERCELL™.Quick answers about timelines, quality, and how to work with Celltheon.

View more answers

Discuss your molecule with our scientists

We don’t workaround — we innovate
and deliver

Explore whether OmniCHO™ is the right expression platform for your specific protein format.

Talk to a Scientist