Duck-Kyun Yoo (Team DK)
Germinal-center (GC) selection is almost always modeled along a single axis — antigen affinity.
Core trick
Folds paired antibody Fv monomers from sequence and reads surface CDR charge as a tolerance axis that transfers across repertoires without retraining.
Description
Germinal-center (GC) selection is almost always modeled along a single axis —
antigen affinity. But B cells are also selected on a second axis: tolerance —
how polyreactive/self-reactive an antibody is. That axis has been hard to use
because it was not measurable at repertoire scale. We make it measurable
directly from structure. We fold paired antibody Fv *monomers* from sequence
(Boltz-2, ESMFold-v1, Protenix) — deliberately avoiding antibody–antigen
complex prediction, which is still unreliable — and read a physicochemical
tolerance axis, surface-displayed CDR charge, off the fold. The score is
folder-invariant, agrees with crystallographic geometry, and transfers without
retraining to an autoimmune (type-1-diabetes) repertoire and to a
tissue-resolved human GC repertoire, where the axis is significantly
constrained in GC B cells. We report the result honestly, including the null
(as a black-box predictor, folded structure does not beat a charge baseline out
of distribution) and the one surviving charge-orthogonal signal. The output is
a mechanistic, reproducible coordinate that single-axis accounts of GC
selection have lacked.
Results (attached again just in case):
https://github.com/dk93js/gc_project, https://drive.google.com/drive/folders/19S4tkwlr4iTsHV2KyxBpoCAm1Cy821XO?usp=sharing
README excerpt
# Two-axis modeling of human germinal center selection via antibody monomer folding Analysis code and input data for the manuscript. We fold paired antibody Fv **monomers** from sequence at repertoire scale and read a physicochemical **tolerance axis** directly off the fold, placing germinal-center (GC) B-cell selection on two axes — antigen-specific **affinity** and **polyreactivity** (self-reactivity / tolerance). We work from the antibody **monomer** on purpose: antibody–antigen *complex* structure prediction is still unreliable, whereas the unbound Fv is well predicted and already carries the tolerance-axis signal. The methodological core is that this axis can be read off a folded monomer as a single, interpretable scalar — **surface-displayed CDR charge** — which is folder-invariant, crystal-validated, and replaces black-box antigen-binding predictors with a mechanistic, reproducible number. ## What the paper shows - **Displayed CDR charge is the structural axis of antibody polyreactivity.** Net sequence charge acts *through* surface-displayed charge; the signal is folder-invariant and sits at the **cross-assay ceiling** of what any polyreactivity score can reach across independent assays. - **The axis transfers to disease repertoires with zero training.** In a type-1-diabetes repertoire, islet-, tetanus- and dsDNA-engaged clones are more positively charged than non-binders, within-donor and donor-robust — a polyreactivity/engagement axis, not autoantigen spec