LASIGE Talks are fortnightly/monthly events to publicize recently distinguished publications or ongoing cutting-edge work by researchers from the research centre, consolidating the scientific culture of the LASIGE community.

Speakers: Francisco Couto (LASIGE/FCUL) and Bakary Badjie (LASIGE)
Date: April 16th, 2025, Wednesday
Where: C6.3.27

Program:
11:45 Francisco Couto
12:05 Bakary Badjie
12:25 Q&A + Break for snacks & coffee

Talk1: Genome-Wide Admixture And Association Study Of Serum Selenium Deficiency To Identify Genetic Variants Indirectly Linked To Selenium Regulation In Brazilian Adults
Speaker: Francisco Couto
Selenium deficiency is a widespread issue linked to impaired antioxidant defense and increased disease risk. Genome-wide association studies (GWAS) can help uncover genetic contributors, yet interpreting SNPs with marginal significance remains challenging. This presentation focuses on Gene Ontology (GO) analysis as a
functional validation tool for GWAS findings. GO terms related to cellular stress and heat response, particularly involving the HIKESHI gene, provide biological support for an indirect genetic link to selenium homeostasis, strengthening the case for SNP relevance despite suggestive p-values.

Talk2: Conducting A Systematic Review: Adversarial Attacks And Defensive Countermeasures On Image Classification Deep Learning For Autonomous Driving
Speaker: Bakary Badjie
The rapid growth of artificial intelligence and Internet of Things technologies has spurred innovation in autonomous driving systems. Image classification models are crucial for decision-making in complex
driving environments. However, these models remain vulnerable to adversarial attacks that can compromise safety in real-time autonomous driving. This systematic review offers an overview of the most recent literature on adversarial attacks and countermeasures on image classification DL models. It highlights the current challenges in mitigating these vulnerabilities. It also presents taxonomies of attacks and countermeasures and outlines recommendations and guidelines for designing effective countermeasures. We suggest interesting
future research directions to enhance model robustness.