eess.IVAug 19, 2026

HiFiC-G: Adapting HiFiC for Hi-C Contact Matrices

Authors: Andre Antonio Straton

Organizations: Faculty of Mathematics and Computer Science, Transilvania University of Bras¸ov (UNITBV)

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

We study whether the loss design of High-Fidelity Generative Image Compression (HiFiC), a GAN-based neural codec originally built for natural photographs, can be adapted to preserve biologically meaningful structure in Hi-C chromatin contact maps under lossy compression. Standard image compression, including HiFiC in its original form, optimizes for human visual perception; but a Hi-C contact map is normally distributed together with its numeric matrix file (.cool/.mcool), which downstream genomic analysis tools consume directly. Aggressive compression that looks acceptable to the eye can nonetheless blur or delete loops and topologically associating domain (TAD) boundaries that these tools depend on. We modify HiFiC's distortion term with a spatially-weighted MSE that up-weights biologically salient regions (loops, TAD boundaries, stripes, compartment structure) and add an insulation-score loss term that directly penalizes loss of TAD boundary sharpness. We describe a three-phase fine-tuning strategy that adapts a pretrained HiFiC checkpoint to the Hi-C domain without catastrophic forgetting. We evaluate the resulting system, HiFiC-G, using both conventional image-quality metrics (PSNR, SSIM) and genomics-domain preservation metrics (loop/TAD/compartment/stripe preservation percentage) across two cell lines. HiFiC-G preserves local structure, meaning stripes and TAD boundaries, substantially better than the metrics alone would suggest, while long-range A/B compartment structure remains poorly preserved; we show this gap tracks genomic scale and is consistent with a specific architectural cause, the fixed-size tiling that both HiFiC-G and the original HiFiC rely on for memory efficiency.

Figures & tables

DatasetRes.CompressionPSNRSSIMLoops %TAD %Compart. %
GM1287825kb411.1 ×\times35.1 dB0.8516.7%32.2%9.9%
GM12878100kb274.0 ×\times35.2 dB0.9325.6%46.8%10.8%
K56225kb466.0 ×\times35.9 dB0.8514.1%62.4%4.1%
K562100kb311.6 ×\times35.6 dB0.9323.7%57.3%5.2%
Table 1: Mean structure preservation by dataset and resolution, statistical detector, corrected source data (n=24 chromosomes per row). Stripe preservation (79–86% across all rows) is omitted from the table for space; see text.
Figure 1: Compression ratio vs. PSNR / SSIM / loop preservation / TAD preservation, all evaluated chromosomes, statistical detector (K562, both resolutions shown; mirrors Figure 4 of Mentzer et al. 2020 ).
Figure 2: SSIM per chromosome at 25kb and 100kb resolution, for GM12878 (top, Sanborn et al. 2015 ) and K562 (bottom, Rao et al. 2014 ). The two cell lines show opposite resolution effects: for GM12878, SSIM is consistently higher at 25kb than at 100kb across every chromosome, while for K562 the reverse holds, SSIM is consistently higher at 100kb. Since both datasets share the same model and evaluation pipeline, this is not an artifact of the compression method itself but reflects a difference in how each dataset’s noise or signal characteristics interact with resolution, consistent with the cross-cell-line discussion in Section 4.3. Neither direction is uniformly better: the choice of resolution trades SSIM against compression ratio and structure preservation differently per cell line, and this figure is included to make that dataset-dependence visible rather than to argue for one resolution over the other.
VariantCompressionPSNRSSIMLoops %TAD %Compart. %Stripes %
hific_med (baseline)416.2 ×\times38.2 dB0.9595.2%82.6%64.9%98.1%
HiFiC-G (fine-tuned)165.7 ×\times28.9 dB0.9594.3%84.6%71.4%98.1%
Table 2: Earlier chr22 ablation-trail result (custom heuristic detector, pre-correction source data), reproduced for comparison against Table 1 .
Figure 3: Example per-chromosome output panel (K562, chr22, 25kb, statistical detector): original matrix, reconstruction, weight map, and detected-structure overlays, generated automatically for every evaluated chromosome. See Section 4.4 for how this figure should and should not be used.
More in this topic

Neural Network Compression

Neural network compression reduces model size, memory, communication, or computation while preserving useful accuracy.

See all 216 papers

Explore similar work

CardsList
Mar 8, 2026

Contact-Guided 3D Genome Structure Generation of E. coli via Diffusion Transformers

In this study, we present a conditional diffusion-transformer framework for generating ensembles of three-dimensional Escherichia coli genome conformations guided by Hi-C contact maps. Instead of producing a single deterministic structure, we formulate genome reconstruction as a conditional generative modeling problem that samples heterogeneous conformations whose ensemble-averaged contacts are consistent with the input Hi-C data. A synthetic dataset is constructed using coarse-grained molecular dynamics simulations to generate chromatin ensembles and corresponding Hi-C maps under circular topology. Our models operate in a latent diffusion setting with a variational autoencoder that preserves per-bin alignment and supports replication-aware representations. Hi-C information is injected through a transformer-based encoder and cross-attention, enforcing a physically interpretable one-way constraint from Hi-C to structure. The model is trained using a flow-matching objective for stable optimization. On held-out ensembles, generated structures reproduce the input Hi-C distance-decay and structural correlation metrics while maintaining substantial conformational diversity, demonstrating the effectiveness of diffusion-based generative modeling for ensemble-level 3D genome reconstruction.

Mingxin Zhang, Xiaofeng Dai, Yu Yao +1

The University of Tokyo · The University of Michigan · Waseda University

GenomicsConditional Diffusion ModelsDiffusion Transformer
AskPDF
May 20, 2026

BlockFormer\textit{BlockFormer} : Transformer-based inference from genomic contact maps

Locating genomic features from genomic contact maps, such as centromere identification from genome-wide chromosome conformation capture techniques, notably Hi-C, can be formulated as an inverse problem: infer one parameter per entity given a map summarizing pairwise interactions through blocks of variable numbers and sizes. In this work, we introduce a data-driven approach that leverages shared structure between these contact maps, such as global alignment between localized patterns, while handling the variability in number and size of chromosomes arising in real-world data. Our approach relies on a transformer architecture capable of handling such variability and a custom simulator to generate abundant, yet computationally cheap synthetic data for training. Applied to the problem of centromere localization, the method recovers genomic positions at or below the map resolution across species with various genome sizes using a coarse-to-fine refinement step when chromosome sizes fall outside the training range. We concentrate our evaluation on Hi-C data, where the block structure is well characterized and ground truth is available.

Eloïse Touron, Pedro L. C. Rodrigues, Julyan Arbel +2

Univ. Grenoble Alpes, Inria CNRS, Grenoble INP, LJK, France · TIMC, Univ. Grenoble Alpes CNRS, Grenoble INP, France

GenomicsInverse Problems
AskPDF
May 16, 2026

Adaptive Fused Prior Transfer for Controllable Generative Image Compression

At very low bitrates, image compression discards fine textures and local structures, while distortion-oriented reconstruction often produces over-smoothed images. Generative codecs synthesize missing details, but existing codebook-based controllable designs generally rely on single-codebook reconstruction priors. We propose Adaptive Fused Prior Transfer for Controllable Generative Image Compression (AFP-GIC), which transfers an image-adaptive fused prior from a frozen pretrained AdaCode model. Encoder-side prior features guide latent formation, while the decoder predicts a compatible fused prior from the compressed representation and control variables, without transmitting the prior itself. A motivating analysis shows that better decoder-side prior alignment tightens a reconstruction-error upper bound and that the fused-prior family includes single-codebook choices as special cases. A single pretrained model supports five evaluated bitrate operating points. Under the unified benchmark, AFP-GIC achieves 18.1% lower decoder latency and uses 31.10 million (20.5%) fewer inference parameters than DC-VIC. Experiments on Kodak, CLIC2020, and DIV2K show competitive PSNR and SSIM, with the clearest naturalness gains in NIQE scores and very-low-bitrate visual comparisons. Code: https://github.com/yifeipet/AFP_GIC.

Yifei Pei, Ying Liu, Nam Ling

Department of Computer Science and Engineering, Santa Clara University, Santa Clara, CA 95053 USA

Learned Image CompressionGenerative Image Compression
AskPDF