Sorry, this website is not ready for mobile preview. Make sure to check on Desktop for now.
AbRefine computational platform

In silico de-risking engineering for whole antibody panels

Profile your whole antibody panel before you commit lab time. AbRefine builds a multiparametric developability profile for every candidate, benchmarks it against more than 100 clinical antibodies, then humanizes and optimizes multiple leads in parallel, reducing lead generation timelines by up to 50%.

Example AbRefine output
Snapshot of 157 antibodies
CDR length
CDR hydrophobicity
CDR positive patch
CDR negative patch
Fv charge symmetry
VH liabilities
VL liabilities
VH humanness
VL humanness
Fv humanness
Antibody ID →
Heatmap of developability scores for 157 antibodies across ten parameters, each cell rated favorable, flag or high risk
Favorable
Flag
High risk
100+
clinical antibodies in the benchmark set
Up to 50%
shorter lead generation timelines
Whole panels
multiple leads humanized in parallel, each with sets of variants
In silico + in vitro
predictions checked against SPR and developability data

Why assess antibody developability in silico first?

Most antibody programs pick a lead on binding data, then discover aggregation, poor stability or charge issues months later in the lab. By then the alternatives in the panel have often been discarded.

AbRefine moves that risk assessment to the start. Large sequence panels are evaluated computationally, so your team progresses the candidates most likely to survive development and spends wet-lab effort only where it counts.

01 Assess

What AbRefine assesses

Antibody sequence, structural modeling and surface property assessment combine into one multiparametric developability profile, run on every sequence in your panel.

Aggregation-prone regions

Hydrophobic surface patches that drive aggregation and poor manufacturability are mapped on the modeled structure.

Chemical modification hotspots

Sequence liabilities such as deamidation and oxidation sites are flagged, with their position and exposure.

Charge characteristics

Positive and negative surface patches and Fv charge are assessed for risks to viscosity, clearance and specificity.

Clinical benchmarking

Every candidate is compared with more than 100 clinical antibodies, so you see where your molecules sit against those that reached the clinic.

Humanness and immunogenicity

Humanness and immunogenicity scores estimate the potential immunogenicity risk of each sequence.

You receive

A ranked candidate list with a multiparametric developability profile per sequence, ready for your next go or no-go decision.

02 Engineer

In silico de-risking engineering

Beyond assessment, AbRefine engineers better variants for multiple candidates at once, each with sets of variants. Humanization runs together with developability profiling, so human sequence content goes up while developability risks go down, in one high-throughput step.

Antibody humanization

Humanized VH and VL variants for multiple leads in parallel, each scored on humanness and developability, with a modeling report.

Affinity maturation

Structure-based modeling and antibody-antigen docking guide which mutations to test.

Liability de-risking

Flagged hotspots and patches are engineered out while binding is preserved.

Full profile for triage

A complete developability profile is generated for every variant, to triage which clones go to the wet lab.

Process

How an AbRefine project works

1

Share your sequences

VH and VL, VHH or scFv sequences from any discovery source, under a confidentiality agreement.

2

Computational triage

Developability profiles, clinical benchmarking and ranking of the full panel.

3

Engineer variants

Humanization, optimization and de-risking of the candidates worth keeping.

4

Validate in the lab

Optional expression, high-throughput SPR on the Carterra LSA XT and a wet-lab developability panel.

Case study

Optimizing a lead antibody for a KRAS-driven oncology program

aSKY Therapeutics, a Baltimore-based therapeutics company, develops antibodies against unique surface antigens on complex tumors associated with mutant KRAS-driven dysregulation. After working with AVS Bio on antibody discovery, aSKY partnered with us in 2026 to optimize its lead antibody for clinical application using AbRefine.

Antigens that only appear on tumor cells under KRAS-driven dysregulation are hard to present in static systems such as yeast display, which slows classic affinity maturation. AbRefine lets the joint team assess affinity-related design choices, humanness and developability in parallel, and focus lab work on the most promising candidates.

“…incredibly impressed with AVS Bio’s committed, transparent, and collaborative way of working with aSKY.”

aSKY Therapeutics, March 2026
AVS Bio scientist running protein purification in the lab
Integrated workflow

Part of an integrated antibody discovery workflow

AbRefine works on sequences from any source, and it is built into AVS Bio’s own discovery platforms: single B cell, phage display, hybridoma and rare hit magnification. Every hit can be ranked from the earliest stage, alongside binding and affinity data.

FAQ

Frequently asked questions

What is in silico antibody engineering?

It uses sequence analysis and structural modeling to assess and improve antibodies before material is made: predicting developability risks such as aggregation, poor stability and unfavorable charge, and designing humanized or optimized variants for lab testing.

How does in silico antibody humanization work?

The non-human framework is replaced with human sequence while the residues that support binding are retained. AbRefine scores each humanized variant for humanness and developability at the same time.

What do I need to send?

Amino acid sequences of your antibodies (VH and VL, VHH or scFv). Antigen information helps for docking and affinity work. We sign a confidentiality agreement before any sequence is shared.

How long does it take?

It can be as fast as 4 weeks after selecting candidates for recombinant expression. Timelines depend on panel size and wet-lab validation; we confirm them in your proposal.

Which antibody formats can AbRefine handle?

AbRefine models the Fv region rather than the whole molecule, so it covers conventional antibodies as well as single-domain VHH and scFv formats.

Scope your project

Find out which of your candidates will make it

Tell us about your panel and we will come back with a scope and timeline.