Batalha Caio M P F, Yu Lei, Zammit Andrea R, Poole Victoria N, Buchman Aron S, Lopes Katia de Paiva, Vialle Ricardo, Abadir Peter, Nidadavolu Lolita, Wyss-Coray Tony, Seyfried Nick T, Wang Yanling, Tasaki Shinya, De Jager Philip L, Iturria-Medina Yasser, Bennett David A
Instituto de Assistência Médica ao Servidor Público Estadual, Sao Paulo, SP, Brazil.
Rush Alzheimer's Disease Center, Rush University Medical Center, Chicago, IL, USA.
bioRxiv. 2025 Aug 18:2025.08.13.670106. doi: 10.1101/2025.08.13.670106.
Here, we define cognitive resilience as slower or faster cognitive decline after we regress out the effects of common brain neuropathologies. Its understanding could provide important insights into the biology underlying cognitive health, enabling the development of more effective strategies to prevent cognitive decline and dementia. However, this requires the development of a practical method to quantify resilience and measure it in living individuals, as well as identifying heterogenous pathways associated with resilience in different individuals. Here, we approach this problem by using a data-driven framework to quantify and characterize molecular signatures underlying cognitive resilience. Using multimodal contrastive trajectory inference (mcTI) on bulk RNA sequencing and tandem mass tag (TMT) proteomic data from 898 post-mortem brain samples from the Religious Orders Study and the Rush Memory and Aging Project (ROSMAP), we derived individual-level molecular pseudotime values reflecting the molecular path from high to low resilience across individuals. Additionally, we identified two distinct molecular subtypes of resilience, each characterized by unique transcriptomic and proteomic signatures, and differing associations with several phenotypes. To translate our brain-derived pseudotime and subtypes to living individuals, we developed prediction models with paired genetics, ante-mortem blood omics, clinical, psychosocial, imaging and device data from the same individuals, demonstrating the potential to predict brain molecular resilience profiles in living persons. Our findings establish a framework for quantifying resilience based on multi-level molecular signatures, identify molecularly distinct resilience subtypes, and demonstrate the feasibility of translating brain-derived molecular profiles to living individuals-laying the groundwork for the development of targeted resilience-promoting interventions in cognitive aging.
在此，我们将认知恢复力定义为在去除常见脑神经病理学影响后认知衰退较慢或较快的情况。对其的理解可为认知健康背后的生物学机制提供重要见解，从而开发出更有效的预防认知衰退和痴呆的策略。然而，这需要开发一种实用方法来量化恢复力并在活体个体中进行测量，同时识别不同个体中与恢复力相关的异质性途径。在此，我们通过使用数据驱动框架来量化和表征认知恢复力背后的分子特征，以解决这一问题。利用来自宗教团体研究和拉什记忆与衰老项目（ROSMAP）的898份尸检脑样本的批量RNA测序和串联质谱标签（TMT）蛋白质组学数据进行多模态对比轨迹推断（mcTI），我们得出了反映个体从高恢复力到低恢复力分子路径的个体水平分子伪时间值。此外，我们识别出了两种不同的恢复力分子亚型，每种亚型都具有独特的转录组和蛋白质组特征，并且与几种表型的关联不同。为了将我们从大脑得出的伪时间和亚型转化到活体个体中，我们利用来自同一批个体的配对遗传学、生前血液组学、临床、心理社会、影像学和设备数据开发了预测模型，证明了预测活体个体脑分子恢复力概况的潜力。我们的研究结果建立了一个基于多层次分子特征量化恢复力的框架，识别出分子上不同的恢复力亚型，并证明了将源自大脑的分子概况转化到活体个体中的可行性，为认知衰老中针对性恢复力促进干预措施的开发奠定了基础。
