New method helps identify young children most at risk of obesity

The study, led by researchers at Penn State University, used data collected from young children to use new statistical methods to establish scoring criteria
.
This study also shows that by collecting comprehensive data over a long period of time, combined with powerful statistical tools, reliable results can be obtained from studies that are orders of magnitude smaller than typical genetic studies

"In the U.
S.
, about 18 percent of children are obese and 6 percent are severely obese," said Sarah Craig, assistant research professor of biology at Penn State
.
"If we can identify children who are most at risk, we may be able to The development of obesity can be prevented in the first place

The study is part of a larger project INSIGHT (Intervention Nurses Keeping Babies on a Healthy Track) coordinated by Penn State Health Milton S.

These risk scores -- known as "polygenic risk scores" because they are based on many genetic positions in the genome -- distill a wealth of genetic information into a single, easy-to-grasp number
.
Typically, these scores contain information from multiple single nucleotide polymorphisms (SNPs), or the positions of DNA letters in the genome in different populations that are linked to metrics of interest (in this case growth rate and obesity) is most relevant

"Previous attempts to score polygenic risk scores for obesity have used genetic information from adults or older children, ranging from 100 million to 2 million SNPs
.
” “Such a high number is challenging for sustained replication and can be very expensive, especially in a clinical setting

The research team used a new statistical technique from the field of functional data analysis to identify the SNPs most associated with obesity, which were then incorporated into the score

The genetic data yielded millions of SNPs that needed to be analyzed, and the team used several techniques to narrow it down to the ones most associated with growth curves and measures of obesity

"We are the first to assess the impact of each SNP individually on obesity-related measures, to eliminate those that are apparently unrelated," said Anna Kenny, a graduate student in the Penn State Statistics study and now a postdoctoral researcher at the University of California
.
"Some studies chose to stop at this point, but we narrowed it down further, looking at all remaining SNPs at the same time, and excluding those that did not appear to have an effect when considered in conjunction with other studies

This process yielded 24 SNPs, which the researchers incorporated into the polygenic risk score
.
Scores based on growth curves have also been shown to correlate with other, more commonly used measures; children with conditional weight gain (change in weight gain in the first 6 months) and infants with rapid weight gain (rapid infant weight gain is a predictors) had higher rates of weight gain
.

The team further narrowed it down to the five most "stable" SNPs -- the ones that had the biggest impact even when perturbing the data
.
From these 5 SNPS, they generated a second score that could serve as an easier option
.

Matthew Reimherr, an associate professor of statistics at Penn State, said: "Although using a score of 24 SNPs is more efficient than using a score of 5 SNPs, we validated that both are useful in measuring obesity risk.
method, we believe either can be used clinically
.
" "A fraction that requires fewer SNPs for identification should make it easier to generate clinically
.
"

Notably, the scores generated in this study also predicted obesity in older children and adults, which the research team validated using publicly available datasets
.
In this study, however, scores from other studies based on information on obesity in adults did not translate to young children
.

"This suggests that the genetic signals we see in early childhood that are associated with obesity persist throughout life," said Ian Paul, professor of pediatrics and public health sciences at Penn State College of Medicine.

"However, as people age, they begin to express other parts of their genetic makeup .

Scores based on early signals appear to be more robust throughout a person's life
.
This highlights the need for more There is a lot of research focusing on identifying risks and preventing obesity in young children, especially in the 'first 1000 days' of pregnancy and the first two years of life
.
"

This study also shows that small studies that deeply characterize individuals and utilize functional data analysis techniques can be a powerful alternative to typical large-scale genetic studies
.

"These techniques could open the door to smaller labs with fewer resources," Craig said
.
"By working carefully and rigorously, collecting longitudinal data from more targeted cohorts, and using powerful statistical techniques, you can still find useful information in a study that is an order of magnitude smaller than a typical GWAS study
.
"

In addition to Craig, Makova, Chiaromonte, Kenney, Reimherr and Paul, the research team included Junli Lin, a Penn State research assistant; Leann Birch, the late professor of food and nutrition at the University of Georgia, who helped Lead INSIGHT; Jennifer Savage, director of the Penn State Center for Childhood Obesity Research and associate professor of nutritional sciences; and Michelle Savage, research technologist and statistician at Penn State's Center for Childhood Obesity Research Marini
.

This work was supported by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK); Penn State Eberly College of Science; Penn State Institute for Computational and Data Sciences; Penn State Huck School of Life Sciences; Pennsylvania Department of Health Additional support was provided by the National Science Foundation for the Use of Tobacco Treatment Fund
.

Journal Reference :

1. Sarah JC Craig, Ana M.
   Kenney, Junli Lin, Ian M.
   Paul, Leann L.
   Birch, Jennifer S.
   Savage, Michele E.
   Marini, Francesca Chiaromonte, Matthew L.
   Reimherr, Kateryna D.
   Makova.
   Constructing a polygenic risk score for childhood obesity using functional data analysis .
   Econometrics and Statistics , 2021; DOI: 10.
   1016/j.
   ecosta.
   2021.
   10.
   014