SPADE: spatial deconvolution for domain specific cell-type estimation.

Understanding gene expression in different cell types within their spatial context is a key goal in genomics research. SPADE (SPAtial DEconvolution), our proposed method, addresses this by integrating spatial patterns into the analysis of cell type composition. This approach uses a combination of single-cell RNA sequencing, spatial transcriptomics, and histological data to accurately estimate the proportions of cell types in various locations. Our analyses of synthetic data have demonstrated SPADE's capability to discern cell type-specific spatial patterns effectively. When applied to real-life datasets, SPADE provides insights into cellular dynamics and the composition of tumor tissues. This enhances our comprehension of complex biological systems and aids in exploring cellular diversity. SPADE represents a significant advancement in deciphering spatial gene expression patterns, offering a powerful tool for the detailed investigation of cell types in spatial transcriptomics.

Semi-reference based cell type deconvolution with application to human metastatic cancers.

Bulk RNA-seq experiments, commonly used to discern gene expression changes across conditions, often neglect critical cell type-specific information due to their focus on average transcript abundance. Recognizing cell type contribution is crucial to understanding phenotype and disease variations. The advent of single-cell RNA sequencing has allowed detailed examination of cellular heterogeneity; however, the cost and analytic caveat prohibits such sequencing for a large number of samples. We introduce a novel deconvolution approach, SECRET, that employs cell type-specific gene expression profiles from single-cell RNA-seq to accurately estimate cell type proportions from bulk RNA-seq data. Notably, SECRET can adapt to scenarios where the cell type present in the bulk data is unrepresented in the reference, thereby offering increased flexibility in reference selection. SECRET has demonstrated superior accuracy compared to existing methods using synthetic data and has identified unknown tissue-specific cell types in real human metastatic cancers. Its versatility makes it broadly applicable across various human cancer studies.

An FPGA implementation of Bayesian inference with spiking neural networks.

Spiking neural networks (SNNs), as brain-inspired neural network models based on spikes, have the advantage of processing information with low complexity and efficient energy consumption. Currently, there is a growing trend to design hardware accelerators for dedicated SNNs to overcome the limitation of running under the traditional von Neumann architecture. Probabilistic sampling is an effective modeling approach for implementing SNNs to simulate the brain to achieve Bayesian inference. However, sampling consumes considerable time. It is highly demanding for specific hardware implementation of SNN sampling models to accelerate inference operations. Hereby, we design a hardware accelerator based on FPGA to speed up the execution of SNN algorithms by parallelization. We use streaming pipelining and array partitioning operations to achieve model operation acceleration with the least possible resource consumption, and combine the Python productivity for Zynq (PYNQ) framework to implement the model migration to the FPGA while increasing the speed of model operations. We verify the functionality and performance of the hardware architecture on the Xilinx Zynq ZCU104. The experimental results show that the hardware accelerator of the SNN sampling model proposed can significantly improve the computing speed while ensuring the accuracy of inference. In addition, Bayesian inference for spiking neural networks through the PYNQ framework can fully optimize the high performance and low power consumption of FPGAs in embedded applications. Taken together, our proposed FPGA implementation of Bayesian inference with SNNs has great potential for a wide range of applications, it can be ideal for implementing complex probabilistic model inference in embedded systems.