Autism Spectrum Disorder (ASD) is neurodevelopmental disorder indicated by difficulties in social interactions and communication, both verbally and non-verbally, as well as the presence of restricted and repetitive behaviors and interests [1]. The prevalence of ASD has been increasing steadily, currently affecting one in 59 children in the United States [2], and the annual national economic burden is estimated to be $265 billion [3].
ASD diagnostics is one focus area of current research. There is general agreement that the earlier ASD is diagnosed, the better the general outcome will be because support services and therapies can start earlier. However, there is no lab test for the diagnosis of ASD and diagnosis is currently performed by observation. The problem with this is that it is difficult to determine if a child is developing normally at a very young age due to the large heterogeneity found in child development but also among children with ASD. This has led to the search for a biomarker or a set of biomarkers that would distinguish a child who will develop ASD from a child who will not.
Research from our lab at Rensselaer Polytechnic Institute has investigated a set of biomarkers to classify between a child with ASD and a typically-developing (TD) peer. This research used metabolites from the Folate-Dependent One-Carbon Metabolism (FOCM) and Transsulfuration (TS) pathways to classify between the two different groups of children with over 95% accuracy [4].
We wanted to see if these metabolites could predict an ASD diagnosis even earlier, so we investigated the possibility of diagnosing child with ASD while they are still in utero. We used a dataset from the MARBLES study at UC Davis that contained data from pregnant mothers who have previously had a child with ASD and the diagnosis of their future child at three years of age [5]. We compared the metabolite measurements at each trimester of the mothers who gave birth to a child diagnosed with ASD at age three with the mothers whose children were developing typically (TD) by age three. We used the Fisher discriminant analysis (FDA) [6] to separate the two groups of mothers and leave-one-out cross-validation to ensure statistical dependence. This approach led to classification errors between 30% and 50% for even the best combination of metabolites. We concluded that using metabolites from the FOCM/TS pathways, we were not able to predict whether the child that a mother is currently pregnant with will be diagnosed with ASD by the age of three.
In addition to the mothers from the MARBLES study, we had access to FOCM/TS metabolite measurements from pregnant mothers who have never had a child with ASD. We decided to compare these mothers to the mothers from the MARBLES study. This analysis can lead to predicting a risk factor of having a child with ASD as mothers who have already had a child with ASD have an 18.7% risk of having another child with ASD, here termed “high risk,” while mothers recruited from the general population have a 1.7% risk, referred to as “low risk.”
We compared the 20 FOCM/TS metabolites collected from these two groups of mothers using the same procedure as before with just the mothers from the MARBLES dataset. From this procedure, we calculated misclassification errors of approximately 10% depending upon which trimester the blood samples were collected. In particular, there was one combination of metabolites that worked well across all three trimesters which included Homocysteine, the ratio of free Cystine to free Cysteine, Glutamylcysteine, free Glutathione, and Nitrotyrosine.
We concluded from this work that metabolites from the FOCM/TS pathways are not able to predict whether or not someone will be diagnosed with ASD while still in utero, but we can predict if the mother is at a higher risk of having a child with ASD while she is pregnant. If it is found that someone has a high risk of having a child with ASD, then approaches can be used to minimize the risk. For example, a recent paper found that folic acid supplementation reduced the risk of having a child with ASD [7]. Additionally, one should also carefully monitor that child’s early development. Although more research is needed to replicate these findings on a larger cohort, this work lays the groundwork for determining ASD risk of the child during pregnancy.
These findings are described in the article entitled Maternal metabolic profile predicts high or low risk of an autism pregnancy outcome, recently published in the journal Research in Autism Spectrum Disorders.
References:
- American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 5th ed. Washington, DC: American Psychiatric Association; 2013.
- J. Baio, “Prevalence of Autism Spectrum Disorder Among Children Aged 8 Years — Autism and Developmental Disabilities Monitoring Network, 11 Sites, United States, 2014,” MMWR Surveill. Summ., vol. 67, 2018.
- J. P. Leigh and J. Du, “Brief Report: Forecasting the Economic Burden of Autism in 2015 and 2025 in the United States,” J. Autism Dev. Disord., vol. 45, no. 12, pp. 4135–4139, Dec. 2015.
- D. P. Howsmon, U. Kruger, S. Melnyk, S. J. James, and J. Hahn, “Classification and adaptive behavior prediction of children with autism spectrum disorder based upon multivariate data analysis of markers of oxidative stress and DNA methylation,” PLOS Comput. Biol., vol. 13, no. 3, p. e1005385, Mar. 2017.
- I. Hertz-Picciotto et al., “Environmental Contributions to ASD Spectrum Disorder: An Introduction to the MARBLES Study,” Environ. Health Perspect., p. In Press, 2018.
- R. Fisher, “The Use of Multiple Measurements in Taxonomic Problems,” Ann. Eugen., vol. 7, no. 2, pp. 179–188, 1936.
- P. Surén et al., “Association Between Maternal Use of Folic Acid Supplements and Risk of Autism Spectrum Disorders in Children,” JAMA, vol. 309, no. 6, pp. 570–577, Feb. 2013.