Empirical Bayes Flow Matching for Continuous Cryo-EM Heterogeneity
Organizations: University of Haifa
Abstract
We address the problem of learning generative priors over high-dimensional latent variables from indirect, noisy observations, with a focus on continuous heterogeneity in cryo-electron microscopy (cryo-EM). In this setting, 3D atomic structures are never directly observed and must be inferred from 2D projection images generated by a complex, non-linear forward operator. We propose Empirical Bayes Flow Matching (EB-FM), a method that learns a flow matching prior directly from observations using a Monte Carlo stochastic approximation Expectation-Maximization framework. EB-FM alternates between an E-step that performs approximate posterior sampling from via guided flow-based generation while maintaining persistent latent estimates through stochastic approximation, and an M-step that updates the flow model using these stabilized latent variables as training targets. This approach eliminates the need for pretraining on clean data and naturally accommodates non-linear forward models. We validate our method on controlled inverse problems using two-moons and MNIST data, and demonstrate its effectiveness on simulated cryo-EM data with known poses, showing that EB-FM can recover continuous distributions of protein conformations directly in atomic-coordinate space.
Figures & tables
| Protein | Metric | GT | EB-FM | VAE | AF3 MSA | cD+gAF3 |
|---|---|---|---|---|---|---|
| 1a62_A | EMD RMSD | – | 1.62 | 1.80 | 1.83 | 1.80 |
| Coverage RMSD | – | 1.37 | 1.65 | 1.56 | 1.58 | |
| Fidelity RMSD | – | 1.27 | 1.46 | 1.18 | 1.01 | |
| Pairwise RMSD GT | 2.40 | 1.58 | 1.20 | 1.23 | 0.77 | |
| RMSF GT | 1.08 | 0.77 | 0.60 | 0.59 | 0.51 | |
| RMSF | – | 0.96 | 0.97 | 0.57 | 0.69 |
Appendix figures & tables5 assets
Supplementary material from the paper’s appendix.
Appendix
| PSNR | SSIM | ||||||
|---|---|---|---|---|---|---|---|
| Obs. | Method | EMD | Fidelity | Coverage | EMD | Fidelity | Coverage |
| 50 | Diffusion | 13.75 | 14.08 | 14.49 | 0.59 | 0.60 | 0.62 |
| EB-FM w/o PL | 13.75 | 14.33 | 14.48 | 0.56 | 0.58 | 0.62 | |
| EB-FM | 15.67 | 16.20 | 16.35 | 0.70 | 0.72 | 0.74 | |
| 100 | Diffusion | 14.72 | 15.36 | 15.42 | 0.68 | 0.70 | 0.71 |
| EB-FM w/o PL | 14.39 | 15.04 | 15.55 | 0.68 | 0.70 | 0.76 | |
| Setting | Diffusion | EB-FM w/o PL | EB-FM |
|---|---|---|---|
| Latent LR | – | – | |
| Guidance steps | 100 | 20 | 20 |
| Persistent latents | No | No | Yes |
| Distributional accuracy | Flexibility | ||||||
|---|---|---|---|---|---|---|---|
| Protein | Method | EMD RMSD Å | Coverage RMSD Å | Fidelity RMSD Å | Pairwise RMSD Å | RMSF Å | RMSF |
| 1a62_A | Ground truth | – | – | – | 2.40 | 1.08 | – |
| EB-FM | 1.62 | 1.37 | 1.27 | 1.58 | 0.77 | 0.96 | |
| EB-FM randomized-init | 1.59 | 1.33 | 1.23 | 1.43 | 0.71 | 0.97 | |
| EB-FM small-error-all | 1.62 | 1.39 | 1.24 | 1.46 | 0.77 | 0.95 | |
| EB-FM p 0.1 error | 1.70 | 1.45 | 1.37 | 1.66 | 0.92 | 0.93 | |
| Parameter | Value |
|---|---|
| Number of picked particles | 10,000 |
| Dose | |
| Image size | px |
| Image pixel size | |
| Magnification | 25,000 |
| Defocus |
Bayesian Inverse Problems
Bayesian inverse problems infer hidden signals or physical parameters from indirect, noisy measurements while quantifying uncertainty.
See all 127 papers