Graph theory as a study means of congenital anomalies associations.

ELIAS, DARIO; COSENTINO VR; GIMENEZ, LUCAS G.; GILI J; RITTLER, MONICA; PAWLUK MS; POLETTA, FERNANDO A; CAMPAÑA H; SANTOS R; RATOWIECKI J; LOPEZ CAMELO JS

Bratislava

Congreso; 46th Annual Meeting of the International Clearinghouse for Birth Defects Surveillance and Research; 2019

International Clearinghouse for Birth Defects Surveillance and Research

Background and Objectives: Since thalidomide and rubella tragedies, surveillance of birth defects has become an public health activity. In order to identify new teratogens, association studies between congenital anomalies have been carried out. To date, two approaches have been used to analyze this associations. The first focuses on a specific anomaly and determines the association degree with other anomalies. The second approach was based on the cases clustering considering each anomaly as a binary variable. The graph theory, offers an integrating approach, allows to analyze the complete set of anomalies and each of them in particular. A graph is constituted by nodes that represent entities and edges that represent the relationship between entities. In recent years this approach has enriched many study areas, for example, proteins interactions and diseases associations. The objective of this work is to analyze the feasibility of this approach to study anomalies complexes and associations. Methods: We used 170,430 cases with at least one anomaly recorded in the Estudio Colaborativo Latino Americano de Malformaciones Congénitas (ECLAMC) between 1967 and 2017. We use the ECLAMC anomaly coding system, which has 207 codes. We used the volume-adjusted Chi-Squared independence test to determine the association strength between anomalies. To partition the graph we used Infomap method. Results: The constructed graph had 98 nodes (68 major anomalies and 30 minor) and 275 edges. Its degree distribution presented a greater adjustment to an Log-Normal distribution than to a Poisson distribution. The graph partition generated 13 anomalies groups. The minor anomalies presented an Eigenvector centrality scores greater than the major anomalies (Wilcoxon p-value 0.0018). Discussion and Conclusions: The congenital anomalies graph obtained differs from random graphs by their degree distribution. The Eigenvector score would show that the minor anomalies were associated more and with highly associated anomalies than the major anomalies, but this result could arise from underreporting of the minor anomalies. The anomalies complexes identified through graph theory would correspond to those found in the literature, such as Patau syndrome. These results would imply that it is feasible to use graph theory for the study of congenital anomalies associations.