Simultaneous optimization of antibody developability & humanness

gilishklar
Oct 30
2 min read

Updated: Nov 5

Scala’s one-round computational humanization generates diverse, fully human antibodies with preserved activity

Why it matters

Classical antibody humanization methods often compromise expression, stability, or function and require lengthy iterative back-mutation cycles. Scala’s CUMAb technology introduces a systematic, computation-driven alternative. By testing thousands of possible human frameworks in silico, CUMAb produces a diverse panel of fully human antibodies predicted to be stable, active, and developable, all in a single design round.

The CUMAb method

CUMAb replaces classical similarity-based humanization with a comprehensive computational workflow. Starting from just an antibody sequence, AI-based predictions generate a structural model of the parental antibody. The mouse CDRs are then computationally grafted onto ~20,000 possible human frameworks. Because these frameworks differ substantially, the process inherently generates a diverse set of fully human antibodies with distinct chemical properties. Atomistic energy calculations are then used to rank the designs and select candidates predicted to combine humanness, stability, and developability.

Problem

A murine monoclonal antibody targeting Quiescin Sulfhydryl Oxidase 1 (QSOX1), an enzyme involved in tumor growth and metastasis, was under development as a cancer therapeutic. Classical humanization approaches failed, yielding antibodies that were not expressible or active, halting the program.

What was done

Using CUMAb, Scala generated 15 fully human variants of the QSOX1 antibody in a single design round. These designs were then experimentally evaluated for expression, stability, activity, and developability.

Results

Design scope: CUMAb generated 15 fully human QSOX1 antibody variants in a single design round.
Expression: 12/15 designs expressed as monomeric IgGs at high yield in mammalian cells.
Binding and activity: 4 variants retained binding affinity within twofold (KD ≤ 2×) of the murine antibody and fully inhibited QSOX1 activity in human plasma.
Stability and developability: Apparent melting temperatures were 70–75 °C; all active variants exhibited low nonspecific binding and minimal self-association relative to a therapeutic IgG benchmark.

Impact

CUMAb succeeded where conventional approaches failed, rescuing the stalled QSOX1 program by delivering multiple fully human antibodies that matched the parental mouse antibody in activity and stability. More broadly, this case demonstrates the general power of CUMAb: a one-shot humanization and optimization method that eliminates iterative back-mutation and delivers a diverse panel of stable, drug-like antibody candidates.

Data highlights

Developability panel: Polyreactivity and self-association plots show that CUMAb antibodies exhibit low nonspecific binding and minimal aggregation compared with clinical IgG benchmarks.

Affinity: Ki plot shows four CUMAb humanized antibodies (haQSOX1.1–1.4) with identical inhibition constants to the parental antibody (maQSOX1).

Activity in human plasma: QSOX1 inhibition assays performed in plasma samples demonstrate that both murine and CUMAb human antibodies maintain equivalent inhibitory activity over time.

References

Tennenhouse A, Khmelnitsky L, Khalaila R, et al. Computational optimization of antibody humanness and stability by systematic energy-based ranking. Nat Biomed Eng. 2024;8(1):30–44. doi:10.1038/s41551-023-01079-1