FACET: Factorized Asymmetric Conditioning for Efficient Transport in High-Fidelity Fluorescence Microscopy Synthesis

Sazan Mahbub; Caleb N. Ellington; Eric P. Xing

cs.LGOct 4, 2026

FACET: Factorized Asymmetric Conditioning for Efficient Transport in High-Fidelity Fluorescence Microscopy Synthesis

Authors: Sazan Mahbub, Caleb N. Ellington, Eric P. Xing

Organizations: Carnegie Mellon University · GenBio AI · Mohamed bin Zayed University of AI

Abstract

Fluorescence microscopy reveals where proteins localize, but only a limited number of proteins can be imaged in the same cell; generating these images from amino-acid sequence and the cell's morphological context enables in silico localization of unimaged proteins. The two conditions, however, play asymmetric roles: morphological context is spatially aligned with the target, whereas sequence is non-spatial and must specify protein-dependent localization within it, with recurring coarse patterns shared across proteins and finer protein-specific variation. Existing generators condition on both jointly, without separating what each explains. We introduce FACET (Factorized Asymmetric Conditioning for Efficient Transport), a probabilistic generative framework that encodes this structure as an explicit inductive bias: sequence semantics are learned from what context leaves unexplained, coarse localization regularities are shared across proteins through a semantic memory, and protein-specific variation is a bounded residual around them. A variance-preserving state projection further lets FACET perform continuous stochastic transport through a pretrained diffusion predictor with minimal parameter overhead. On held-out proteins, FACET improves spatial overlap by 34.3% on the Human Protein Atlas and 14.0% on OpenCell over a backbone-matched baseline, and reduces FID by 27.2% and 46.5%, respectively, with 75% fewer network evaluations. It also substantially improves protein-association structure recovery and yields better-calibrated predictions, while detailed ablations show complementary contributions from its design choices. These results identify factorized asymmetric conditioning, rather than generator capacity alone, as a key lever for high-fidelity, efficient, and biologically meaningful cellular image synthesis.

Figures & tables

Panel of Figure 1 — Figure 1: Probabilistic graphical model for FACET .

Figure 2: Representative held-out HPA examples: morphology context, target, and model outputs. CELL-E2 uses nucleus-only context; CELL-Diff and FACET use all three context channels.

Method	NFE $\downarrow$	MI $\uparrow$	MS-SSIM $\uparrow$	Haralick $\uparrow$	PSNR $\uparrow$	$\mathrm{IoU}_{\mathrm{cont}}$ $\uparrow$	FID $\downarrow$
CELL-E2	–	0.4041	0.1370	0.5644	6.36	0.1755	166.4
CELL-Diff	100	0.5355	0.3471	0.5182	15.77	0.2478	45.60
FACET	25	0.5481	0.3651	0.5854	16.15	0.3328	33.20

Table 1: Comparison on the HPA test set. Higher is better for all metrics except NFE and FID. Best results are shown in bold.

Context	System	MI $\uparrow$	MS-SSIM $\uparrow$	Haralick $\uparrow$	PSNR $\uparrow$	$\mathrm{IoU}_{\mathrm{cont}}$ $\uparrow$	FID $\downarrow$
Nucleus	CELL-Diff	0.3091	0.2777	0.4932	15.32	0.2047	51.10
Nucleus	FACET	0.2519	0.1712	0.5010	14.66	0.2768	52.12
Nucleus + ER	CELL-Diff	0.4562	0.3154	0.4898	15.52	0.2182	60.00
Nucleus + ER	FACET	0.5120	0.3413	0.5726	15.96	0.3268	33.83
Nucleus + MT	CELL-Diff	0.5097	0.3404	0.5196	15.60	0.2435	47.60
Nucleus + MT	FACET	0.5226	0.3524	0.5796	16.03	0.3266	33.24

Table 2: Performance across cellular-context compositions on HPA. All systems use the same proteins, sequence inputs, and sampling protocol per composition. Bold marks the better value within each composition.

Method	NFE $\downarrow$	MI $\uparrow$	MS-SSIM $\uparrow$	Haralick $\uparrow$	PSNR $\uparrow$	IoU cont $\uparrow$	FID $\downarrow$
CELL-E2	–	0.1649	0.1020	0.4542	11.31	0.1779	245.8
CELL-Diff	100	0.2732	0.4078	0.3663	13.20	0.3722	68.90
FACET (HPA-adapted)	25	0.2753	0.3653	0.5177	13.85	0.4222	43.33
FACET (OpenCell-adapted)	25	0.2998	0.3893	0.5199	13.75	0.4244	36.83

Table 3: Comparison on the OpenCell test set (nucleus-only context). All systems except CELL-E2 share the OpenCell-trained CELL-Diff backbone; HPA-adapted transfers the FACET modules from HPA without refitting, whereas OpenCell-adapted fits them on OpenCell. Best results are in bold.

System	STRING score threshold	Pairs (assoc./null)	Pearson $\Delta$ $\uparrow$	$p_{\mathrm{assoc.\ vs.\ null}}$
Baseline	0.85	47/2659	+0.1005	0.298
FACET	0.85	47/2659	+0.1875	$1.3\times 10^{-3}$
Baseline	0.70	65/2641	+0.0954	0.170
FACET	0.70	65/2641	+0.1718	$4.5\times 10^{-4}$
Baseline	0.50	112/2594	+0.0922	0.042
FACET	0.50	112/2594	+0.1369	$2.5\times 10^{-4}$

Table 4: Spatially resolved protein-association pattern recovery under a localization-matched evaluation adapted from ( Sun et al., 2026 ) . The baseline denotes CELL-Diff. Pearson similarity is computed after removing the per-cell mean image.

\Delta

measures the separation between STRING-associated and matched non-associated pairs; positive values indicate stronger recovery.

Figure 3: Quality–cost comparison on OpenCell.

FACET configuration	$\varepsilon$ -adapter	Memory-conditioned semantic pathway	Backbone LoRA (base velocity $v_{L}$ )	Hierarchical head (velocity correction $\Delta v$ )	$\mathrm{IoU}_{\mathrm{cont}}$ $\uparrow$
FACET (full)	✓	✓	✓	✓	0.3328
Memory-conditioned + LoRA	✓	✓	✓	–	0.3236
Memory-conditioned + hier. head	✓	✓	–	✓	0.3052
Memory-conditioned	✓	✓	–	–	0.2826
$\varepsilon$ -adaptation-only	✓	–	–	–	0.2581

Table 5: Controlled FACET configuration ablation on the HPA benchmark. Each configuration is evaluated on the same held-out proteins using identical cellular context, sampler, and preprocessing. Checkmarks indicate active pathways. “hier.” represents “hierarchical”.

Configuration	OpenCell adaptation	HPA module initialization	OpenCell-fit memory	$\mathrm{IoU}_{\mathrm{cont}}$ $\uparrow$
FACET, HPA-adapted	–	$\checkmark$	–	0.4222
FACET, OpenCell-adapted, cold start	$\checkmark$	–	$\checkmark$	0.4244
Warm start, inherited HPA memory	$\checkmark$	$\checkmark$	–	0.4155
Warm start, refitted OpenCell memory	$\checkmark$	$\checkmark$	$\checkmark$	0.4233

Table 6: HPA-module transfer and OpenCell adaptation ablation. A checkmark indicates that the corresponding component is active. Note that, for all the configurations, the

\varepsilon

-predictor backbone, VAE encoder/decoder, and protein LM are adopted from the OpenCell checkpoint by Zheng and Huang (2025) .

Configuration	Learned memory	Memory replaced at inference	$\mathrm{IoU}_{\mathrm{cont}}$ $\uparrow$
FACET, unmodified memory	$\checkmark$	–	0.3328
FACET, uniform prototype mean	–	$\checkmark$	0.2820
FACET, zero memory	–	$\checkmark$	0.3014

Table 7: Ablation of inference-time memory on HPA. A checkmark indicates that the corresponding memory configuration is active.

Sampler	NFE	FID $\downarrow$	$\mathrm{IoU}_{\mathrm{cont}}$ $\uparrow$
SDE	10	41.88	0.4301
SDE	25	36.83	0.4244
SDE	50	36.18	0.4214
ODE (Heun’s)	25	36.60	0.4205
ODE (Heun’s)	50	36.17	0.4187

Table 8: OpenCell sampling results across solvers and network function evaluations (NFE).

Method	ECE $\downarrow$	MCE $\downarrow$	Brier $\downarrow$
CELL-Diff	0.21501	0.24886	0.23372
FACET	0.05442	0.14850	0.19915

Table 9: Predictive calibration on HPA. One temperature per model, fit on a protein-disjoint validation split and held fixed on test. Foreground probabilities come from 16 predictive samples per input.

Appendix figures & tables8 assets

Supplementary material from the paper’s appendix.

Appendix

Setting	Value
Semantic memory vocabulary	$K=9$ fixed UniProt localization categories
Residual representation	128 dimensions
Interpolant strength	$\kappa=0.5$
Correction head	2 residual blocks
LoRA adaptation	rank $16$
Inference budget	25 NFE

Appendix

Table 10: Selected implementation settings for the FACET instantiation used in the reported experiments; component-level ablations are reported separately.

Panel of Figure 4 — Figure 4: FACET inference overview. The protein sequence directly conditions the pretrained $\varepsilon$ -predictor and queries the semantic memory to obtain $m_{S}$ , while a sequence-conditioned residual refines it into $z_{S}$ . The memory-conditioned $\varepsilon$ -adapter forms the base velocity $v_{L}$ , and context-first hierarchical modulation produces the correction $\Delta v$ . Their composition $v_{\Theta}=v_{L}+g(\tau)\Delta v$ is iteratively integrated through stochastic latent transport, yielding the final latent estimate for decoding by the frozen VAE. Training-only paths are omitted; see Algorithm 2 for details.

Module	Parameters	Deployed
$\varepsilon$ -adapter	1,318,533	✓
Prototype prior $p_{\eta}$	923,145	✓
Hierarchical correction head	4,918,340	✓
LoRA (rank 16)	2,406,400	✓
Continuous prior $p_{\psi}$	262,785	✓
Total deployed	9,829,203	–

Appendix

Table 11: Parameter counts for the trainable FACET modules. Deployed modules are used during generation; posterior modules are used only during training.

Figure 5: Cellular-context ablation. The protein sequence is held fixed while the cellular-context input is removed or modified. The examples illustrate context-conditioned changes in the generated spatial phenotype; they are qualitative evidence rather than a causal ablation beyond the quantitative comparison reported in the main text.

Figure 6: Same-protein generation across HPA cellular contexts. The sequence input is fixed across contexts, allowing context-conditioned changes in the generated spatial phenotype to be visualized.

Figure 7: Additional HPA generations across multiple proteins and cellular contexts. The examples provide broader coverage of the conditional-generation setting beyond the main-text qualitative comparison.

Figure 8: Additional qualitative results on OpenCell. The figure uses the same method names and visualization conventions as the main OpenCell results and complements the quantitative comparison. Here “CELL-Diff (fine-tuned)” represents the CELL-Diff checkpoint fine-tuned on OpenCell data, the same one reported in the main text and the source ( Zheng and Huang, 2025 ) .

Figure 9: Additional qualitative results on HPA. CELL-E2 uses nucleus-only context, whereas CELL-Diff and FACET use the full three-channel context. These examples extend the main-text qualitative comparison using the same evaluation setting, method names, and visualization conventions.

Explore similar work

ProHiFlo: Hierarchical Flow Matching with Functional Guidance for De Novo Protein Generation
Jun 3, 2026Chuanzhen Wang, Meade Cleti, Pete JanoFlow Matching 3D Molecular Generation
3Tongji University · 1Arizona State University · University of Wisconsin-Madison
Asymmetric Flow Models
May 13, 2026Hansheng Chen, Jan Ackermann, Minseo Kim +2Flow Matching T2I Generation
Stanford University
MicroZoom: Structure-Preserving Detail Synthesis at Extreme Scale
Jul 27, 2026Huy Huynh, Jingwei Ma, Brian Curless +2Image Generation Texture Synthesis
University of Washington, USA

Jun 3, 2026, cs.LGJ/K move · Enter open · S save

ProHiFlo: Hierarchical Flow Matching with Functional Guidance for De Novo Protein Generation

Chuanzhen Wang, Meade Cleti, Pete Jano

3Tongji University · 1Arizona State University · University of Wisconsin-Madison

Flow Matching 3D Molecular Generation Protein Design Conditional Generative Modeling

Open paper AskPDF arXiv

De novo protein generation has transformative potential in therapeutic design, enzyme engineering, and synthetic biology. While diffusion-based and flow matching approaches have achieved progress, they typically operate at single resolution and lack mechanisms for incorporating functional constraints. We introduce ProHiFlo, a hierarchical flow matching framework with three innovations: (1) coarse-to-fine generation that models backbone geometry before refining to all-atom coordinates, reducing computational cost while maintaining accuracy; (2) functional guidance leveraging pretrained predictors to steer generation toward desired properties without retraining; (3) adaptive SE(3)-equivariant architecture for efficient multi-scale processing. Experiments on unconditional generation, motif scaffolding, and functional design demonstrate state-ofthe-art performance while requiring 4 fewer sampling steps. On enzyme active site scaffolding, ProHiFlo achieves 58.9% success rate compared to 41.2% for RFDiffusion.

Abstract

Explore similar work

ProHiFlo: Hierarchical Flow Matching with Functional Guidance for De Novo Protein Generation

Asymmetric Flow Models

MicroZoom: Structure-Preserving Detail Synthesis at Extreme Scale