Deep Learning Scientist Data Science Yinxiu Zhan Lab

1000scholars; Milano; On-site; EUR 41,000 - 50,000; Full time

Job summary

IEO - European Institute of Oncology in Milan, Italy, seeks a Deep Learning Scientist to design CNN and Vision Transformer models that operate directly on RNA-seq data to detect expressed mutations. The role interfaces with bioinformatics and ML teams to benchmark performance and integrate calls into predictive pipelines.

The project emphasizes RNA-driven cancer genomics and immuno-oncology, with opportunities to contribute to scientific publications and software documentation.

Qualifications

• Strong Python programming and data analysis skills.
• Experience with PyTorch or TensorFlow and CNN/Transformer models.
• Familiarity with RNA-Seq / DNA-Seq workflows and variant calling pipelines.

Responsibilities

• Develop and optimize deep learning architectures for RNA-seq–based variant calling.
• Adapt and extend CNN- and ViT-based frameworks for RNA-specific mutation detection.
• Benchmark RNA-based variant calls against DNA-seq ground truth datasets.

Skills

Python

NumPy / Pandas / SciPy

Linux HPC

PyTorch / TensorFlow

CNNs / Transformers

RNA-seq data representations

Education

PhD or Master’s in a quantitative field

Tools

STAR

GATK

samtools

bcftools

Job description

Deep Learning Scientist Data Science Yinxiu Zhan Lab

IEO - European Institute of Oncology

Milan, Italy

€ 45.000 per year gross

About the Project

Accurate RNA-based detection of genetic variants has the potential to streamline molecular profiling by extracting multiple layers of information from a single experimental technique. Unlike approaches that require separate assays for DNA variant calling and transcriptomic readouts, RNA sequencing can simultaneously capture gene expression profiles and evidence of expressed mutations, features that are particularly relevant when studying tumor biology and treatment response.

Within the PRIME project, the Fellow will focus on designing, implementing, and benchmarking a hybrid CNN-Vision Transformer (ViT) framework that operates directly on RNA-seq-derived data to enable robust detection of expressed mutations. The work will include model development and optimization, definition of evaluation strategies and benchmarks, and systematic comparison against baseline approaches to quantify performance, generalizability, and practical utility. This effort contributes to PRIME’s broader goal of improving prediction of response to immune checkpoint inhibitors through RNA-driven computational methods.

Key Responsibilities

• Develop and optimize deep learning architectures for RNA-seq–based variant calling
• Adapt and extend DeepVariant-like frameworks for RNA-specific mutation detection
• Implement CNN and Vision Transformer models for local and global sequencing feature extraction
• Benchmark RNA-based variant calls against matched DNA-seq ground truth datasets
• Design validation pipelines and performance metrics (precision, recall, F1-score)
• Collaborate with bioinformatics and machine learning teams to integrate variant calls into downstream predictive models
• Contribute to scientific publications and technical documentation

Required Competencies

Programming & Data Analysis

• Advanced proficiency in Python
• Experience with scientific computing libraries (numpy, pandas, scipy)
• Familiarity with Linux-based HPC environments

Machine Learning & Deep Learning

• Strong experience with deep learning frameworks (PyTorch or TensorFlow)
• Solid understanding of CNNs and transformer-based architectures
• Experience working with sequencing data representations

Genomics & Bioinformatics

• Experience in next-generation sequencing (RNA-seq, DNA-seq)
• Experience with read alignment, variant calling, and quality control pipelines
• Familiarity with tools such as STAR, GATK, samtools, bcftools

Soft Skills

• Strong analytical and problem-solving skills
• Ability to work independently and within interdisciplinary teams
• Clear communication skills and attention to reproducibility

Desirable Qualifications

• Experience with DeepVariant or similar variant calling frameworks
• Familiarity with FFPE sequencing data
• Background in cancer genomics or immuno-oncology

Educational Requirements

• PhD or Master’s degree in Mathematics, Physics, Bioinformatics, Computational Biology, Computer Science, or related fields
• Demonstrated research experience in deep learning

1 day left