| Sumario: | OBJECTIVE: Preterm birth is a serious issue in the medical world that can affect the family, especially the mother, both mentally and physically. Even babies need to go through so many short- and long-term health complications and need to suffer throughout their lives. Utmost, it can lead to the loss of life. Early identification of preterm birth can help doctors improve the critical condition or may help their efforts to stop that premature delivery. Hence, research and studies toward the early identification of preterm delivery are very crucial. Different data, such as images, blood samples, clinical data, and omic-scale data, can be used in the identification of various diseases and other disease-related studies. Even though other data can do a reasonable job in this research, molecular data could do a better job as they carry most information related to a human. METHODS: This study compares the performance difference between transcriptome and epigenome data in the prediction of preterm birth. As it is very difficult to get these data from an unborn baby, data obtained from the mother are used. It starts with feature selection using mutual information. Twenty features are selected from each dataset and used separately with six different classification algorithms. RESULTS: Comparing the performance between models shows that transcriptome data give a better prediction accuracy using the K-nearest neighbor classification algorithm, with the accuracy value of 0.96 (±0.07). CONCLUSIONS: This study shows that transcriptome data from the mother can be used in the prediction of preterm birth.
|
|---|