THE ISABS FUTURE SCIENTIST AWARD PRESENTATIONS
Presentation number: FSA 01
Abstract number: 169-ISABS-2026
INVESTIGATION OF GENOTOXICITY AND CYTOTOXICITY OF UV NAIL LAMP RADIATION
Cvitković Petra1, Tabak Klara1
1III. Gymnasium, Split, Croatia
petra.cvitkovic2508@gmail.com; amaretto.klara@gmail.com
The aim of the research was to determine toxicity (genotoxicity and cytotoxicity) of radiation emitted by UV, LED, and combined UV/LED nail lamps using the Allium cepa test. The experiment investigated the effects of different radiation types and exposure durations on the growth and genetic material of onion root tips. Germinated seeds of Allium cepa L. were exposed to UV (36W), LED (5W), and UV/LED (54W) lamps for intervals of 2.5, 5, 7.5, and 10 minutes. Root length was measured to assess cytotoxicity, while the mitotic index, chromosomal aberrations, and micronuclei were analyzed via microscopy to evaluate genotoxic effect. The results showed that UV lamp radiation caused significant growth inhibition of the roots, particularly at the 10-minute exposure mark (p=0.02). Combined UV/LED radiation resulted in growth inhibition and morphological changes during longer exposure intervals, whereas LED radiation showed no significant impact on root length under the experimental conditions. While the mitotic index did not show statistically significant deviations, a significant increase in chromosomal aberrations and micronuclei was observed in the UV and UV/LED treatment groups. Specifically, UV radiation for 7.5 minutes induced aberrations in 70.84% of dividing cells (p=0.00002). These findings confirm the genotoxic potential of UV and UV/LED nail lamps. Given that an accompanying survey revealed that over 80% of users apply no skin protection during treatments, further education on the risks and preventive measures is essential.
Keywords: genotoxicity, cytotoxicity, UV radiation
Presentation number: FSA 02
Abstract number: 171-ISABS-2026
USING THE ALPHAGENOME AI MODEL TO INTERPRET A GENETIC VARIANT LINKED TO FAMILIAL HYPERCHOLESTEROLAEMIA
Kokić Martina1
1Geodetic School Zagreb, Zagreb, Croatia
martina.kokic11@gmail.com
To test whether the AlphaGenome AI model can quickly and accurately analyse a known harmful genetic variant in the LDLR gene – a gene linked to Familial Hypercholesterolaemia (FH), a hereditary condition causing dangerously high cholesterol – and to explore whether this kind of AI analysis could help doctors make faster diagnostic decisions. AlphaGenome (developed by Google DeepMind, published in Nature, 2026) was used through its Python programming interface in Google Colab. A 1-megabase stretch of DNA around the variant position was analysed across four biological readouts: RNA-seq, DNase-seq, CHIP-TF and splice site signals. A ‘local effect score’ was calculated to measure how strongly the variant disrupts each readout near its exact position. Statistical significance was tested using the Mann-Whitney U test. A gene ranking tool (score_variant) was also used to compare the LDLR gene against all 57 other genes in the region. All four readouts showed signs of harmful impact. The effect scores were: RNA-seq 7.8σ, DNase 72.1σ, CHIP-TF 51.5σ, and splice sites 477.6σ. The Mann-Whitney U test gave U= 0 and p ≈ 0, meaning the result is highly statistically significant. LDLR ranked 6th most affected out of 57 genes. AlphaGenome produced strong, consistent evidence that the variant rs121908027 in the LDLR gene is harmful – in under five minutes. Standard genetic testing for the same conclusion takes four to eight weeks. These results suggest that AI tools like AlphaGenome could help doctors reach faster diagnoses, as long as they are used responsibly and always reviewed by qualified medical professionals.
Keywords: familial hypercholesterolaemia, AlphaGenome, artificial intelligence, genetic variant interpretation
Presentation number: FSA 03
Abstract number: 170-ISABS-2026
APPLICATION OF ARTIFICIAL INTELLIGENCE IN THE DIAGNOSIS AND TREATMENT OF RARE GENETIC DISEASES
Prša Krunić Irina1
1X. Gymnasium Ivan Supek, Zagreb, Croatia
prsairina89@gmail.com
To date, approximately 8,000 rare diseases have been identified. Although each affects a small number of individuals, collectively they represent a major global health challenge, impacting nearly 400 million people worldwide (about 5.7% of the population). Around 80% have a genetic origin, with symptoms often emerging in early childhood. Despite this, diagnosis often takes years or even decades, and about 95% of rare diseases remain without approved treatments. In this context, this study explores the potential of artificial intelligence as a transformative tool for improving the diagnosis and treatment of rare diseases. This study is based on a review of scientific literature, primarily utilizing the PubMed database along with additional relevant sources, focusing on genetics, bioinformatics, and the clinical application of artificial intelligence. A review of the literature indicates that artificial intelligence enables rapid and highly accurate analysis of complex genetic and clinical data, identifying pathogenic variants among hundreds of thousands of alterations generated through whole-genome sequencing. Machine learning algorithms can detect subtle patterns often missed by traditional methods. Advanced AI systems such as DeepRare demonstrate high diagnostic accuracy, frequently outperforming conventional approaches while significantly reducing time to diagnosis. By integrating genetic, clinical, and phenotypic data, artificial intelligence accelerates diagnosis and supports the development of targeted, personalized therapies. Overall, it is redefining rare disease care by enabling earlier detection, more precise treatment, and advancing precision medicine.
Keywords: artificial intelligence, machine learning, rare genetic diseases, next-generation sequencing
Presentation number: FSA 04
Abstract number: 168-ISABS-2026
EFFECT OF HYDROGEN PEROXIDE ON Na-K ATPASE ABUNDANCEIN NEURO2A CELLS
Vodanović Ana Karla1
1XV. Gymnasium, Zagreb, Croatia
anakarla.vodanovic@gmail.com
The aim of this study is to investigate the effect of different concentrations of hydrogen peroxide on the abundance of sodium-potassium ATPase in a mouse neural cell line, Neuro2a, in order to further determine to what extent oxidative stress induces changes in gene expression in addition to its known effects on enzyme activity. Neuro2a cells were exposed to a range of H2O2 concentrations (10 nM to 10 mM) for 15 minutes to model both physiological and oxidative stress conditions. Following overnight incubation, cells were lysed and protein extracts were analysed using Western blotting. Na-K ATPase abundance was quantified and normalised to beta-actin. Data was analysed using one-way ANOVA to assess statistical significance between treatment groups. No statistically significant differences in Na-K ATPase abundance were observed across the tested H2O2 concentrations, as ��=0.82. High variability between replicates and a low coefficient of determination, ��2=0.0259, indicated the absence of a clear relationship between H2O2 concentration and protein abundance. No consistent trend was observed across treatment groups. These findings suggest that short-term oxidative stress does not significantly affect Na-K ATPase abundance in Neuro2a cells. The results support the idea that hydrogen peroxide primarily influences enzyme activity through rapid posttranslational mechanisms rather than inducing changes in gene expression under acute conditions. Further studies incorporating longer exposure times and complementary functional assays are required to determine whether sustained oxidative stress leads to transcriptional regulation of Na-K ATPase.
Keyword: Neuro2a cells, Na-K ATPase, hydrogen peroxide effect, beta-actin
Presentation number: FSA 05
Abstract number: 167-ISABS-2026
ARTIFICIAL INTELLIGENCE IN PRECISION ONCOLOGY: ADVANCING CANCER DIAGNOSIS AND TREATMENT
Weiser Elena1
1I. Gymnasium, Varaždin, Croatia
elena.weiser@skole.hr
The aim of this work is to explore how artificial intelligence (AI) can transform precision oncology. It examines the biological complexity of cancer and compares traditional therapies and AI-driven tools in diagnosis, treatment selection, and patient outcomes. This study is based on a review of scientific literature, with the PubMed database used as the primary source. The selected literature includes studies focusing on artificial intelligence, cancer biology, precision medicine, and targeted therapy in oncology. Analysis shows that AI facilitates precision medicine by tailoring treatment to the specific genetic profile of a patient and the unique heterogeneity of their tumor. Furthermore, by utilizing AI-assisted predictions for immunotherapy and targeted therapies, clinicians can identify the most effective drugs for individual patients. These AI applications demonstrate significant potential in minimizing treatment side effects, detecting tumors at earlier stages, and ultimately improving patient survival rates. While cancer remains a complex disease requiring highly individualized approaches, AI significantly improves the precision of data analysis and clinical decision-making. The integration of AI into oncology promises a revolution in medicine through advancements such as AI-assisted drug discovery and real-time monitoring of tumor evolution. In the new era of precision medicine, artificial intelligence will not replace oncologists, but oncologists who use artificial intelligence will likely replace those who do not.
Keywords: artificial intelligence, precision oncology, cancer genomics, machine learning, personalized medicine
Published: June 16th, 2026.
Copyright: © 2026 Authors of section FSA. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
