Study Uncovers Biomarkers for Low Muscle Mass, Offering Insights for Personalized Prevention

By Dr. Chinta SidharthanReviewed by Lily Ramsey, LLMAug 15 2024

Sarcopenia, or a deterioration in the mass, quality, and strength of muscles, results from age-related complications or underlying comorbidities. While it is independently associated with decreased physical abilities, sarcopenia also contributes significantly to increased injury and disease risk.

In a recent study published in Scientific Reports, a team of researchers used machine learning models to analyze computed tomography (CT) images of the chest and a proteomics approach to determine blood protein biomarkers that could together be used to assess the low muscle mass risk.

​​​​​​​Study: Proteomics and machine learning in the prediction and explanation of low pectoralis muscle area. Image Credit: Toa55/Shutterstock.com

Background

The loss of muscle mass and the deterioration of muscle strength and quality are associated with a significant decrease in the quality of life and an increased risk of mortality.

While sarcopenia could result naturally with age, it is also an indicator of other comorbidities and increases the risk of injury and disease. Identifying the risk of sarcopenia and implementing interventions before the occurrence of other adverse events is essential.

CT images of the chest are used to measure the pectoralis muscle area (PMA), one of the most commonly used methods of assessing sarcopenia.

Assessments of muscle quality based on PMA measurements have been used previously to predict the worsening of respiratory conditions.

Various biomarkers and clinical factors, such as demographic factors, inflammatory biomarkers, and comorbidities, have also been linked to low muscle mass. However, there is a shortage of research on validating PMA measurements as an effective predictor of low muscle mass and identifying the drivers of low muscle mass.

About the Study

In the present study, the researchers used data from the COPDgene study, which conducted longitudinal observations among current and former smokers to examine the genetic epidemiology of the chronic obstructive pulmonary disease (COPD).

The data was collected in two phases, one at baseline and the second after five years, and consisted of extensive questionnaires, CT scans of the chest, and protein biomarker measurements from blood samples.

Axial CT images of the suprasternal notch and the aortic arch were used to derive PMA measurements. With the PMA of control participants who had never smoked as a reference, low PMA was defined as values falling below the 25^th percentile and was stratified based on sex.

The biomarker measurements were compared with and without low PMA baseline measurements between current and former smokers to identify potential protein biomarkers.

The researchers then built deep learning models to predict the probability of developing low PMA in five years. These models were also used to test the effectiveness of the protein biomarkers in predicting low PMA development.

One of the models used only clinical measurements such as weight and height and demographic characteristics, including sex and age, to predict the probability of low PMA development.

Another model used only the protein biomarker data for the prediction, while the third model incorporated both clinical measurements and protein biomarker data for low PMA development prediction.

The contributions of the biomarker and clinical predictors in group and individual PMA development risk were calculated, and the relevance of the selected protein biomarkers in the change in PMA in five years was also assessed.

Major Findings

The results showed that the model that incorporated clinical and demographic data and blood protein biomarkers effectively predicted the probability of low PMA development in current and past smokers. The study also identified eight protein biomarkers as important predictors of low PMA.

The eight biomarkers were CDON or cell adhesion molecule-related/down-regulated by oncogenes, epidermal growth factor-containing fibulin-like extracellular matrix protein 1 (EFEMP1), growth differentiation factor 15 (GDF15), histone acetyltransferase type B catalytic subunit (Hat1), lymphotoxin α1/ β1, neurexophilin-1 (NXPH1), SPARC or secreted protein acidic and rich in cysteine, and vascular cell adhesion protein 1 (VCAM-1).

The markers identified in the study were diverse in their functions, ranging from histone modification to regulation of leukocyte migration.

Furthermore, the role of some of the markers, such as GDF15, in the development of PMA has been validated by previous studies.

In contrast, associations between markers such as CDON and human muscle mass remain unknown. However, murine models have shown that CDON is involved in skeletal muscle development.

Conclusions

Overall, the findings demonstrated that machine learning models could potentially be used to predict the development of low PMA using clinical measurements and protein biomarker information.

This would enable the early implementation of dietary, pharmacological, and exercise-based therapies to prevent adverse outcomes and lower the risk of disability and disease.