Building on the. success of diffusion models in visual generation, flow-based models reemerge as another prominent family of generative models that have achieved competitive or better performance in terms of both visual quality and inference speed. By learning the velocity field through flow-matching, flow-based models tend to produce a straighter sampling trajectory, which is advantageous during the sampling process. However, unlike diffusion models for which fast samplers are well-developed, efficient sampling of flow-based generative models has been rarely explored. In this paper, we propose a framework called FlowTurbo to accelerate the sampling of flow-based models while still enhancing the sampling quality. Our primary observation is that the velocity predictor's outputs in the flow-based models will become stable during the sampling, enabling the estimation of velocity via a lightweight velocity refiner. Additionally, we introduce several techniques including a pseudo corrector and sample-aware compilation to further reduce inference time. Since FlowTurbo does not change the multi-step sampling paradigm, it can be effectively applied for various tasks such as image editing, inpainting, etc. Besides, we propose a new multi-stage refinement technique that is designed to reduce the inference costs with large flow-based image generation models. Specifically, the multi-stage refinement split the whole generation procedure on different resolutions, forming a coarse-to-fine text-to-image pipeline. We further adopt a stage-aware deployment strategy that can maximize the inference speed in terms of both latency and throughput. By integrating FlowTurbo into different flow-based models, we obtain an acceleration ratio of 53.1%$\sim$∼58.3% on class-conditional generation and 29.8%$\sim$∼38.5% on text-to-image generation. Notably, FlowTurbo reaches an FID of 2.12 on ImageNet with 100 (ms/img) and FID of 3.93 with 38 (ms/img), achieving the real-time image generation and establishing the new state-of-the-art. Equipped with the recent SD 3.5 Large, we achieved FID of 28.05 with a speed improvement of around 50% on NVIDIA 3090 GPU.

Building on the success of diffusion models in visual generation, flow-based models reemerge as another prominent family of generative models that have achieved competitive or better performance in terms of both visual quality and inference speed. By learning the velocity field through flow-matching, flow-based models tend to produce a straighter sampling trajectory, which is advantageous during the sampling process. However, unlike diffusion models for which fast samplers are well-developed, efficient sampling of flow-based generative models has been rarely explored. In this paper, we propose a framework called FlowTurbo to accelerate the sampling of flow-based models while still enhancing the sampling quality. Our primary observation is that the velocity predictor's outputs in the flow-based models will become stable during the sampling, enabling the estimation of velocity via a lightweight velocity refiner. Additionally, we introduce several techniques including a pseudo corrector and sample-aware compilation to further reduce inference time. Since FlowTurbo does not change the multi-step sampling paradigm, it can be effectively applied for various tasks such as image editing, inpainting, etc. By integrating FlowTurbo into different flow-based models, we obtain an acceleration ratio of 53.1\%$\sim$58.3\% on class-conditional generation and 29.8\%$\sim$38.5\% on text-to-image generation. Notably, FlowTurbo reaches an FID of 2.12 on ImageNet with 100 (ms / img) and FID of 3.93 with 38 (ms / img), achieving the real-time image generation and establishing the new state-of-the-art. Code is available at https://github.com/shiml20/FlowTurbo.

Phase separation is an important mechanism that mediates the spatial distribution of proteins in different cellular compartments. While phase-separated proteins share certain sequence characteristics, including intrinsically disordered regions (IDRs) and prion-like domains, such characteristics are insufficient for making accurate predictions; thus, a proteome-wide understanding of phase separation is currently lacking. Here, we define phase-separated proteomes based on the systematic analysis of immunofluorescence images of 12 073 proteins in the Human Protein Atlas. The analysis of these proteins reveals that phase-separated candidate proteins exhibit higher IDR contents, higher mean net charge and lower hydropathy and prefer to bind to RNA. Kinases and transcription factors are also enriched among these candidate proteins. Strikingly, both phase-separated kinases and phase-separated transcription factors display significantly reduced substrate specificity. Our work provides the first global view of the phase-separated proteome and suggests that the spatial proximity resulting from phase separation reduces the requirement for motif specificity and expands the repertoire of substrates. The source code and data are available at https://github.com/cheneyyu/deepphase.

MOTIVATION: Synonymous codon usage bias has been shown to be correlated with many genomic features among different organisms. However, the biological significance of codon bias with respect to gene function and genome organization remains unclear. RESULTS: Guanine and cytosine content at the third codon position (GC3) could be used as a good indicator of codon bias. Here, we used relative GC3 bias values to compare the strength of GC3 bias of genes in human and mouse. We reported, for the first time, that GC3-rich and GC3-poor gene products might have distinct sub-cellular spatial distributions. Moreover, we extended the view of genomic gene domains and identified conserved GC3 biased gene domains along chromosomes. Our results indicated that similar GC3 biased genes might be co-translated in specific spatial regions to share local translational machineries, and that GC3 could be involved in the organization of genome architecture. AVAILABILITY AND IMPLEMENTATION: Source code is available upon request from the authors. CONTACT: [email protected] or [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.