Protein synthesis is dynamically regulated to control cell growth, differentiation, and stress responses. Recent single-cell sequencing methods can map ribosome positions on individual transcripts1–4, but cannot capture the global translational states that coordinate protein synthesis across the transcriptome. In contrast, methods that measure the global translational landscape, such as polysome profiling and cryogenic electron tomography5, lack either single-cell resolution or throughput. Here we introduce SCISSOR (Single-Cell Inference of Structural States of Ribosomes), a strategy that infers global translation activity in individual cells from the differential protection of ribosomal RNA (rRNA) against nuclease digestion. By integrating these protection signatures with the structure of the ribosome, SCISSOR resolves multiple ribosomal states and quantifies their abundance across thousands of individual cells. Applying SCISSOR reveals systematic variation in global translation across the cell cycle in human cells, as well as during the differentiation of murine intestinal stem cells into distinct epithelial lineages. These findings uncover principles of global translational regulation that are invisible to transcriptomic or ribosome-profiling assays, establishing a framework for studying global translation control at single-cell resolution.
Euan Joly-Smith, Michael VanInsberghe, Kseniia Sarieva et al.· bioRxiv· 0 citations
GPUs are increasingly used for analytical query processing, but developing GPU-based database engines that achieve the peak performance of the underlying hardware requires substantial research and engineering effort. A recent line of work argues that query processing should be synthesized, not engineered. In this scenario, instead of tuning a general-purpose engine to fit a workload, a large language model (LLM) generates code specialized to one query, one dataset, and one machine, thereby achieving an order-of-magnitude improvement in performance. This thesis, however, has so far been tested only on CPUs. In this work, we revisit the synthesize-versus-engineer debate for GPU analytics by answering three questions: (i) how good is synthesized GPU code?, (ii) why is it faster than engineered engines?, and (iii) how much of its advantage can be transferred back into a single, performance-portable engine? To answer the first question, we present SHADB, an LLM-based synthesis framework that generates optimized CUDA or HIP kernels using an automated, profile-guided optimization loop. Using SHADB, we show that the synthesized code approaches the memory-bandwidth ceiling and outperforms a state-of-the-art JIT-compiled GPU database engine (HeavyDB) by 7.4$\times$ on SSB SF100. To answer the second question, we decompose this performance gap and systematically classify optimizations as generalizable or workload-specific. Finally, to answer the third question, we integrate these generalizable optimizations into SYCLDB, a performance-portable engine written entirely in the open SYCL programming model. Using optimized SYCLDB, we show that it is possible to substantially bridge the gap to synthesized code (within 1.27$\times$ total execution time) while retaining workload-level generality and hardware-level performance portability.
I. D. Kabadzhov, Eugenio Marinelli, Raja Appuswamy· 1 citation
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
