Aartik Sarma, MD
Assistant Adjunct Professor
I am a physician-scientist at UCSF and affiliate faculty in the UCSF/UC Berkeley Joint Program in Computational Precision Health. I am an attending physician in the medical and neurological ICUs at UCSF Parnassus and my NHLBI-funded research focuses on applying a wide range of computational omics methods to understand critical illness syndromes and respiratory infections.
The long-term objective of my translational research program is to identify precision treatments for acute respiratory diseases and critical illness by understanding causes of disease heterogeneity and selecting relevant preclinical disease models. My current research includes:
1) Transcriptomic analyses of acute respiratory diseases: I have used RNASeq to characterize the heterogeneous respiratory biology of patients with ARDS. I identified distinct features of respiratory tract gene expression in patients with COVID-19 ARDS (Sarma Nat Comms 2021) and studied the effect of dexamethasone on pulmonary and systemic inflammation in COVID-19 (Neyton*, Patel*, Sarma* Nat Comms 2024).
My research program is currently supported by a NHLBI K23 to study the evolution and resolution of ARDS molecular phenotypes defined by plasma biomarkers and clinical variables (Calfee Lancet Resp Med 2014). I have used bulk and single cell RNASeq from the lower respiratory tract in three cohorts and found increased interferon-stimulated gene expression and T cell activation in the lower respiratory tract in patients with a higher acuity ARDS phenotype (Sarma medRxiv 2022).
In addition, I am developing methods to predict individualized treatment effects in ICU trials and received a grant from the American Thoracic Society to study patients enrolled in two randomized trials of mesenchymal stem cells in ARDS patients to identify biomarkers associated with a treatment response.
2) Translational significance of preclinical lung injury models: Acute respiratory diseases are highly heterogeneous, and it is highly unlikely that preclinical disease models are equally relevant to all patients. I have studied the effects of inhaled nicotine in a murine model of influenza (Maishan*, Sarma* Front Immuno 2023) and the effects of corticosteroids on S. pneumoniae infections in mice and mechanically ventilated patients (Taenaka*, Wick*, Sarma* Crit Care 2024). I am now applying machine learning methods to identify shared features between preclinical model transcriptomes and transcriptomes from patients with acute respiratory diseases.
3) Algorithmic bias in the ICU: Medical algorithms are invaluable tools in the ICU, but can measurement errors introduce subtle biases into patient care with large effects on outcomes. I identified a systematic bias in the formulas used to set ventilator tidal volumes that may contribute to excess mortality in patients with ARDS.
4) Non-invasive sampling of the lower respiratory tract: Biological specimens from the respiratory tract, including sputum, tracheal aspirates, and bronchoalveolar lavage, contain diluted samples of biomarkers from epithelial lining fluid in the lung. While several methods have been proposed to normalize biomarker concentrations from these samples, none adequately accounts for dilution effects, and there is no consensus on the optimal methods to study these biospecimens (Bain, Bastarache, Sarma, et al. AnnalsATS 2024). I am co-chairing an American Thoracic Society Workshop on to establish research standards for non-invasive sampling of the lower respiratory tract.