Machine learning-based proteomics profiling of ALS identifies downregulation of RPS29 that maintains protein homeostasis and STMN2 level
Wei Xu, Zhipeng Guo, Yian Guan, Shihui Lv, Xue Gao, Wenchen Luo, Tian‐Lin Cheng, Zhicheng Shao, Bangbao Tao, Tao Wang, Zhixin Qiu
Abstract
Amyotrophic lateral sclerosis (ALS) is a devastating motor neuron disease. The molecular understanding of ALS is hampered by the lack of experimental models recapitulating disease heterogeneity and analytical framework integrating multi-omics datasets. Here, we developed a pipeline integrating machine learning and consensus clustering to analyze a large-scale dataset of patient-derived motor neuron models from Answer ALS. Compared to the transcriptome, proteomic profiling closely correlates with ALS pathology, which is interrogated to identify 110 proteomics-based biomarkers (Proteomics Markers for ALS 110, PMA110). Functional enrichment highlights dysregulation of ALS pathways, including protein translation and neuronal function. By integrating ALS subtype-specific proteins with patient postmortem proteomics, we found that RPS29 was consistently downregulated in ALS models and patient motor neurons. RPS29 is required for neuronal viability by maintaining ribosome profiling and accurate translation, and suppressing pathological translation. RPS29 downregulation suppresses translation of STMN2, an essential protein for motor neurons, in iPSC-derived motor neurons. Taken together, this study provides a robust framework for ALS proteomics, identifies RPS29 as a quality controller of protein translation, and presents a translational mechanism for STMN2 maintenance in ALS.