OUR METHODS · STATISTICAL MACHINE LEARNING

Cryo-ET Reconstruction

Missing-wedge restoration is, at heart, an inverse problem — self-supervised, ground-truth-free, solved with statistical machine learning.

no ground truth less missing wedge

Strengths & limits

What they do well: the reconstructions land closer to the real structure than IsoNet or DeepDeWedge, they put back detail the missing wedge had erased, and they need no ground-truth labels — CryoWGEN even tells you where it is least sure. The catch: CryoGEN gives you a single answer with no sense of how certain it is, and the part it fills into the wedge is inferred, not actually measured.

Articles in this base 5 articles

The four methods — lineage & comparison

CryoGEN-I
MAP point estimate
Tool MAP + energy prior, P_Y-proxy adversarial learning
Limit → Single point estimate; no uncertainty
CryoGEN-II
OT distribution matching
Tool WAE: global distribution match, pure optimization
Limit → Still one deterministic reconstruction per observation
CryoWGEN-I
Entropic · Monte-Carlo
Tool EVIA: entropic reg → Boltzmann posterior, MC sampling
Limit → MC sampling may be insufficiently precise
CryoWGEN-II
Entropic · iterative Langevin
Tool EVIA: SGLD iterative sampling, more faithful to the posterior
CryoGEN · one curve — a single answer (point estimate)
CryoGEN-IMAP — a single answer; the GAN-style energy carries bias, and it is overconfident
CryoGEN-IIglobal distribution matching (optimal transport) — a more stable single answer, but still GAN-family bias
CryoWGEN · a family of curves — answers with uncertainty (a distribution)
CryoWGEN-IMonte-Carlo — entropic-smoothed energy, a family closer to the truth (coarser)
CryoWGEN-IILangevin — the same smooth energy; the most faithful sampling, the tightest band
true structureCryoGEN (one)CryoWGEN (a family)

The true structure is two peaks (grey dashed). The missing wedge makes the gap between them ambiguous — and the four methods answer it differently. CryoGEN gives one definite answer, but it learns a GAN-style energy surface that carries bias: CryoGEN-I (MAP) deviates most and is overconfident; CryoGEN-II uses optimal transport to stabilize training and match the overall distribution, so it deviates less — but it is still a single deterministic answer. CryoWGEN switches to entropic regularization (EVIA), whose energy surface is smoother — its reconstructions sit closer to the truth, and instead of one answer it returns a family: CryoWGEN-I by Monte-Carlo (coarser, widely spread), CryoWGEN-II by Langevin / SGLD (the most faithful sampling, the tightest band). The width of that band is the missing-wedge uncertainty, made readable. Drag the slider — the more is missing, the more ambiguous the gap.

Papers & results