Resumo:
Autism is an age- and sex-related lifelong neurodevelopmental condition characterized pri marily by persistent deficits in core domains such as social communication. It is estimated
that ≈ 2% of children have some ASD trait. The autism etiology is mainly due to inherited
genetic factors (>80%). The importance of early diagnosis and interventions motivated
several studies involving groups at high risk for ASD, those with a greater predisposition
to the disorder. Such studies are characterized by evaluating some characteristics of the
individual itself or the family members of diagnosed individuals, mainly aiming to predict
a future diagnosis or recurrence rates. One of the primary goals of Artificial Intelligence
is to create artificial agents capable of intelligent behaviors, such as prediction problems.
Prediction problems usually involve reasoning with uncertainty due to some information
deficiency, in which the data may be imprecise or incorrect. Such solutions may seek the
application of probabilistic methods to construct inference models. In this thesis, we will
discuss the development of probabilistic networks capable of estimating the risk of autism
among the family members given some evidence (e.g., other family members with ASD).
In particular, the main novel contributions of this thesis are as follows: the proposal of
some estimates regarding parents with ASD generating children with ASD; the highlight ing regarding the decrease in the ASD prevalence sex ratio among males and females
when genetic factors are taken into account; the corroboration and quantification of past
evidence that the clustering of ASD in families is primarily due to genetic factors; the
computation of some estimates regarding the risk of ASD for parents, grandparents, and
siblings; an estimate regarding the number of ASD cases in a family sufficient to attribute
the ASD occurrences to the genetic inheritance; the assessment of some estimates for
males and females individuals given evidence in grandparents, aunts-or-uncles, nieces-or nephews and cousins; and the proposition of some estimates indicating risk ranges for
ASD by genetic similarity.
