Steinhubl Steven R, Sekaric Jadranka, Gendy Maged, Guo Huaijian, Ward Matthew P, Goergen Craig J, Anderson Jennifer L, Amin Sarwat, Wilson Damen, Paramithiotis Eustache, Wegerich Stephan
Purdue University, W, Lafayette, IN, USA.
Prolaio Health, Scottsdale, AZ, USA.
Commun Med (Lond). 2025 Apr 13;5(1):115. doi: 10.1038/s43856-025-00840-8.
Effective response to vaccination requires activation of the innate immune system, triggering the synthesis of inflammatory cytokines. The degree of subjective symptoms related to this, referred to as reactogenicity, may predict their eventual immune response. However, the subjective nature of these symptoms is influenced by the nocebo effect, making it difficult to accurately quantify a person's physiologic response. The use of wearable sensors allows for the identification of objective evidence of physiologic changes a person experiences following vaccination, but as these changes are subtle, they can only be detected when an individual's pre-vaccination normal variability is considered.
We use a wearable torso sensor patch and a machine learning method of similarity-based modeling (SBM) to create a physiologic digital twin for 88 people receiving 104 COVID vaccine doses. By using each individual's pre-vaccine digital twin, we are able to effectively control for expected physiologic variations unique to that individual, leaving only vaccine-induced differences. We use these individualized differences between the pre- and post-vaccine period to develop a multivariate digital biomarker for objectively measuring the degree and duration of physiologic changes each individual experiences following vaccination.
Here we show that the multivariate digital biomarker better predicted systemic reactogenicity than any one physiologic data type and correlated with vaccine-induced changes in humoral and cellular immunity in a 20-person subset.
A digital biomarker is capable of objectively identifying an individual's unique response to vaccination and could play a future role in personalizing vaccine regimens.
对疫苗接种的有效反应需要激活先天免疫系统，触发炎性细胞因子的合成。与此相关的主观症状程度，即反应原性，可能预测其最终的免疫反应。然而，这些症状的主观性受到反安慰剂效应的影响，使得难以准确量化一个人的生理反应。可穿戴传感器的使用能够识别个体接种疫苗后所经历的生理变化的客观证据，但由于这些变化很细微，只有在考虑个体接种前的正常变异性时才能检测到。
我们使用可穿戴的躯干传感器贴片和基于相似性建模（SBM）的机器学习方法，为88名接种104剂新冠疫苗的人创建生理数字孪生模型。通过使用每个个体接种前的数字孪生模型，我们能够有效控制该个体特有的预期生理变化，只留下疫苗诱导的差异。我们利用接种前后的这些个体差异来开发一种多变量数字生物标志物，以客观测量每个个体接种疫苗后所经历的生理变化的程度和持续时间。
我们在此表明，多变量数字生物标志物比任何一种生理数据类型都能更好地预测全身反应原性，并且在一个20人的子集中与疫苗诱导的体液和细胞免疫变化相关。
数字生物标志物能够客观识别个体对疫苗接种的独特反应，并可能在未来的疫苗接种方案个性化中发挥作用。
