Fixstars is a provider of software and hardware for research labs, medical professionals, and biotech businesses to answer biological questions. We are assisting researchers all around the world in producing real results from their biological data with our reliable and reproducible analysis procedures and well-annotated solutions.
Unlock insights from the industry's largest archive of linked, longitudinal real-world computing solutions
Engage our engineering experts to architect computing systems for your bio research.
Leverage our extensive bioinformatics knowledge and great skills to create and implement custom apps across all areas to meet your requirements.
Scientists, artists, and engineers require massive parallel computational capability. Fixstars provides GPU-optimized virtual machines that take advantage of the capabilities of CUDA, Tensor, and RT cores to run complicated processing, deep learning, and ray tracing tasks.
NoRMCorre is an algorithm widely used among scientists, and most commonly used in calcium imaging for neuroscience. However, it is also one of the most time-consuming pipelines used for neuroanalysis. We port and optimize NoRMCorre to GPU, and our GPU version will run twice as fast as the original CPU version of NoRMCorre in a rigid mode data set. This service will improve the overall lead time.
You concentrate on your key talents, while we worry about the hardware that enables them. Fixstars GPU machines lower the entry barrier for complicated use cases like machine learning and AI.
Fixstars GPUs are NVIDIA A100, which are now regarded as among the best on the market. These GPUs support any use case connected with parallel processing, deep learning, or ray tracing thanks to CUDA, Tensor, and RT cores in each unit.
Is one GPU card insufficient for your anticipated workloads? It's not a problem. Fixstars GPU plans include up to four cards per instance, depending on the amount of horsepower required.
A highly multiplexed method for precisely finding numerous RNA molecules within cells and tissues at the same time.
The in-situ sequencing (ISS) facility offers spatially resolved gene expression data at the subcellular level for panels of genes.
We speed up the software used for in-situ sequencing by identifying bottlenecks and optimizing the computer code. Whether it's GPU clusters, high-speed data storage, or one of our other services, we'll employ whatever is best for your applications.
Overcoming Microscope Limitations
Expansion microscopy (ExM) is a method that physically magnifies tissues in an isotropic manner, allowing for super-resolution imaging using diffraction-limited microscopes and a huge field of view.
As a result, ExM is well-positioned to integrate molecular content and cellular morphology, with the spatial precision required to resolve individual biological building blocks as well as the scale and accessibility required to deploy over extended 3-D objects such as tissues and organs.
Learn More about ExM Studio
The ExM Studio is a software package for Expansion Microscopy. It accelerates the software for the two primary components of the imaging process: deconvolution and stitching.
With the GPU acceleration you get the same high-quality results as before, but much faster! Using high-end NVIDIA GPU cards you achieve astonishing deconvolution results in seconds. The GPU mode also has the distinctive brick-splitting capability, allowing you to deconvolve very huge files on the GPU,with limited video -RAM. Furthermore, the algorithm can accurately compensate for spherical aberration in the event of a refractive index mismatch.
GPU acceleration runs a small piece of the application code that requires substantial computing time on the GPU, while the rest of the code runs on the CPU. Because the GPU is made up of hundreds or even thousands of distinct processing units (cores), it is incredibly efficient and quick at processing computationally intensive code.
We offer a scalable hybrid CPU-GPU image stitching system that works with huge image sets at nearly interactive speeds. Both picture sizes and the number of CPU cores and GPU cards in a machine grow well with our method.
With this implementation, coarse-grain parallelism is utilized. The computation is organized into a pipeline architecture that utilizes both CPU and GPU resources and overlaps computation with data movement. The translation of each tile is estimated during the initial phase of deployment.
The method outperforms our optimized non-pipeline GPU implementation by almost 10 times and exhibits near-linear speedup with increasing CPU thread count and GPU count.
The tiles are joined together into a mosaic in the second stage using the calculated translations. However, because they create an overly restricted set of equations, these translations cannot be applied directly. By treating the system as a graph and gradually merging strongly-connected components, our method overcomes this constraint.
We gauge the strength of the connections using the cross-correlation values. By reducing uncertainty, this optimization improves stitching precision.
Validated Strong Performance and Functionality
All-Flash NAS is a fast, cost effective all-flash storage device for 3D microscopes. It stores, writes, and reads microscope data at I/O rates of 23-25Gbps for reads and 5Gbps for random writes (Lightsheet LiveScan needs a minimum of at least 2 Gbps for writing).
Microscope image reading & writing and I/O will be much faster and cost effective with this accelerated technology. Images can be transferred back and forth between microscope, software and storage systems at a fast rate.
UCLA Dr. Dong's laboratory introduced a microscope all-flash NAS based on QNAP where 25Gbps connection is available between 3D microscopes and the AFA.
Fixstars tuned up the microscope all-flash NAS for Lightsheet workloads and it reaches 23Gbps reads and 5Gbps small file writes with the cost-efficient read-intensive SSD arrays. By our solution, they can enjoy full bandwidth for image reads in data analysis and enough random writes for Lightsheet's LiveScan workload.
Read the full case study
