Banca de QUALIFICAÇÃO: GABRIELLE GERMEK COELHO SANTOS

Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
STUDENT : GABRIELLE GERMEK COELHO SANTOS
DATE: 31/07/2026
TIME: 14:00
LOCAL: Google Meet
TITLE:

Characterization of Rare Genetic Variants in Brazilian Populations Using Sequencing Data

KEY WORDS:

genomics; variant; brazil; pathogenic; prediction.

PAGES: 53
BIG AREA: Ciências Biológicas
AREA: Biologia Geral
SUMMARY:

Genomic variants are alterations in the DNA sequence that occur at the individual or population level. Understanding their implications, mechanisms, and origin is essential for advancing precision medicine strategies. The Brazilian population, shaped by a history of extensive miscegenation, harbors unique variants and phenotypes. Yet, Brazilian genomic data remain underrepresented in major international databases, largely due to the population’s genetic complexity and the limited functional interpretation of available data. This gap hinders the progress of genomics-based studies in Brazil. This study aimed to identify and characterize rare germline genetic variants in Brazilian individuals using next-generation sequencing (NGS) data, as well as to reclassify the pathogenicity of previously reported variants using deep learning-based approaches implemented in DTreePred, a new mobile application developed by Gomes et al. (2025) for assessing variant pathogenicity. As a case study, we reanalyzed the variants identified in a Brazilian cohort described by Amorim et al. (2025), consisting of 17 samples. The analysis followed a standard protocol, beginning with quality control of the reads using Trimmomatic and alignment to the GRCh38/hg38 reference genome using BWA-MEM, followed by sorting, duplicate marking, and indexing with Samtools and Picard Tools. Variant calling was performed using DeepVariant in WES mode and annotation was done using a custom Perl script, integrating information from public databases. Finally, pathogenicity prediction was performed with DTreePred, which applies machine learning models to classifying variants, and other complementary tools, such as CADD, MutPred2 and EVE. Driver/passenger prediction was conducted with CScape and Cancer Genome Interpreter (CGI) tools. The investigation resulted in the detection of 4 new variants that hadn't been described in most known genetic databases. When analyzing the functional profile of the variants, DTreePred classified all of them as pathogenic, unlike other repositories and tools, like ClinVar and SIFT, which didn’t provide a classification or described as of conflicting interpretation or VUS (Variant of Uncertain Significance). CScape and CGI demonstrated significant results, contributing to the pathogenic profile of the variants and possible cancer driver roles. These findings emphasize the potential of deep learning tools like DTreePred to enhance variant interpretation in underrepresented populations. By uncovering novel candidate driver variants, this study expands the catalog of Brazilian genomic diversity and establishes a framework for future functional studies in precision medicine.

COMMITTEE MEMBERS:
Presidente - 2170415 - JORGE ESTEFANO SANTANA DE SOUZA
Interno - 1046922 - LEONARDO CAPISTRANO FERREIRA