Machine learning model identifies clock genes linked to seasonal affective disorder

Key takeaways:

• A machine learning model predicted seasonal affective disorder and depression.
• The model identified sex-specific clock genes that were associated with seasonal affective disorder and depression.

WASHINGTON — A machine learning model accurately predicted seasonal affective disorder and depressive symptoms and identified sex-specific clock genes associated with these symptoms, according to a poster presented here.

Thanh Dang, an undergraduate research assistant at Colgate University in Hamilton, New York, and colleagues recruited 273 students and faculty members from their university to participate in the study. Participants self-reported seasonal affective disorder (SAD), depression, sleep disturbance, diurnal preference, chronotype, socioeconomic status and gender.

The researchers collected genetic data to identify nine single nucleotide polymorphisms (SNPs). They used machine learning modeling to evaluate the associations between participant genotypes and phenotypes and SAD and depressive symptoms.

Dang presented the data at the Anxiety and Depression Association of America Conference.

The models predicted seasonality with 62.5% accuracy, which was greater than the baseline random chance prediction accuracy of 56.5%. They also predicted female seasonality with 69.4% accuracy and male seasonality with 68.2% accuracy.

In female participants, CRY2/PER3C and CRY2/PER3-VNTR were significant risk factors for seasonality and SAD; in male participants, CLOCK3111/ZBTB20 and PER2/PER3B were significant risk factors.

Clinical risk factors for seasonality and SAD in female participants were anxiety, eveningness and older age, according to the poster.

“We were excited to discover the direct effects of CRY1 and ZBTB20 on seasonal mood symptoms as well as depressive symptoms,” Dang told Healio. “In conjunction with previous reports that CRY1 and ZBTB20 are associated with major depressive disorder, our results suggest that these two clock-related genes may also have direct effects on mood pathways involved in seasonal affective disorder.”

These two SNPs may be biomarkers for seasonality and depression, and could help in the development of SAD treatments, according to Dang.

“To address the relatively small and predominantly Caucasian population in this study, we are conducting mood disorder research on a larger and more diverse dataset, the UK Biobank,” she told Healio. “We are hopeful that findings from more representative and expansive genetic data will have greater generalizability and meaningful clinical implications. We are also studying SAD and circadian rhythms in a population of Karen (Burmese) refugees who have relocated to Upstate New York to better understand how diverse populations have adapted to varied photoperiodic environments at different latitudes.”