PhD Student in Artificial Intelligence: Interpretable Vector Symbolic Architectures for Genome Interpretation

The Institut de Génétique Moléculaire de Montpellier (IGMM) is seeking a motivated PhD student to join the 'AI for Genome Interpretation' lab directed by Daniele Raimondi. This interdisciplinary project focuses on utilizing Vector Symbolic Architectures (VSA), or Hyperdimensional Computing, to model genotype-phenotype relationships. The successful candidate will develop multi-scale representations of genomic data to overcome the limitations of current deep learning models regarding interpretability and data scarcity. The research involves prototyping methods on yeast datasets before applying them to human exome sequencing data for disease risk prediction. The position offers an opportunity to work at the intersection of machine learning, bioinformatics, and clinical genetics.

• Develop multi-scale VSA representations for genomic sequencing data using binary, bipolar, ternary, and continuous encodings.
• Design and implement supervised and self-supervised learning methods for genomic hypervectors.
• Benchmark VSA-based models against conventional machine learning and deep learning architectures.
• Create interpretable decoding methods to quantify the contribution of variants, genes, and pathways to predictions.
• Apply developed methodologies to yeast genotype-phenotype prediction and human IBD disease-risk analysis.
• Analyze the biological relevance of model-discovered associations.
• Publish research findings in high-impact international journals and conferences.