Below the list of proposed projects for the XL CQB PhD cycle.
Design and Synthesis of Cell-Type Specific Promoters for Gene Therapy Using Artificial Intelligence
This project aims to boost development of gene therapy applications through the design and synthesis of cell-type specific promoters using advanced artificial intelligence techniques. By leveraging single-cell transcriptomic data from the Human Cell Atlas, you will develop a powerful AI-driven platform to create short promoters tailored for distinct cell types or sets of cell types. The project will utilize transformers and large language models (LLMs) to learn gene expression patterns and features, followed by fine-tuning these models with reinforcement learning (RL) to optimize promoter sequences for targeted gene expression. The designed promoters will be experimentally validated to ensure specificity and efficacy in therapeutic gene expression. This interdisciplinary research offers a unique opportunity to integrate computational biology, genomics, and machine learning, providing graduate students with hands-on experience in cutting-edge AI methodologies and their applications in gene therapy.
PI: Prof. Diego di Bernardo (This email address is being protected from spambots. You need JavaScript enabled to view it.)
High-throughput screening of synthetic and natural vesicles in cells and machine learning model
High-throughput screening (HTS) of synthetic and natural vesicles in cells combined with machine learning (ML) model development is an advanced approach to discover and optimize vesicle-based systems for various applications, such as drug delivery, diagnostics, and therapeutics. This project will integrate the design of nanoparticles and machine learning with high-throughput experimental platforms to identify nano-bio interactions. The aim is to better understand the role of nanotechnology in biological environments and potentially accelerate the translation of these nanoplatforms to clinical use through precise design. Potential applications include cancer therapy, immunotherapy, and diagnostics.
Prof. Enza Torino (This email address is being protected from spambots. You need JavaScript enabled to view it.)
This PhD project is finalised to design a comprehensive framework for information fusion of biological data with heterogeneous data gathered from different sensors. This fusion will enable advanced AI-based data modelling to analyse and predict trends and anomalies in the agritech domain. By leveraging cutting-edge artificial intelligence (AI) techniques, we integrate heterogeneous data, uncover insights, and forecast trends to support the decision-making processes, enabling the detection of anomalies such as pest infestations, disease outbreaks, and equipment malfunctions.
This project will address the critical need for robust methodologies to integrate and analyse multi-source data, ultimately contributing to enhanced agricultural practices and food security.
The Mediterranean basin is highly vulnerable to climate change, qualifying it as a hotspot.
Durum wheat (Triticum turgidum ssp. durum, genome AABB) is extensively cultivated in this region, playing a significant role in food security, economy, culture, and environment. However, stressors such as drought, salinity, and poor nutrient availability adversely impact wheat yield and quality. These stresses lead to the overproduction of reactive oxygen species (ROS), resulting in oxidative stress. Effective ROS detoxification is critical for maintaining plant productivity, necessitating the regulation of antioxidant enzymes. Polyphenol oxidase (PPO) enzymes, which act as ROS scavengers, are crucial in this process.
This project aims to identify and characterize the polyphenol oxidase (PPO) gene family members in durum wheat (T. turgidum ssp. durum) and bread wheat (T. aestivum) using in silico methods. Genome sequences and gene annotations will be obtained from public repositories. HMM-profiles generated from custom multiple sequence alignments and retrieved from the Pfam database will be utilized to scan these genomes, identifying known and potentially novel PPO genes through the HMMER software package. Downstream in silico characterization will follow established workflows.
Following identification and characterization, targeted RNA-sequencing via Ion AmpliSeq™ technology will be employed to capture and enrich selected PPO genes. Total RNA from wheat accessions will be isolated, assessed for quality, and used to create sequencing libraries. Sequenced reads will be analyzed to generate a comprehensive PPO gene expression atlas.
Additionally, Ion AmpliSeq reads will be aligned to reference genomes for variant calling. SnpEff and PredictSNP will predict the impact of detected variants on protein function. Sanger sequencing will confirm SNP calls and discover new allelic variants within PPO genes. The deliverables of this project will include a complete list of PPO gene family members, a gene expression atlas, and a catalog of nucleotide variants impacting PPO genes.
Develop novel NGS pipelines for population genomic screenings of rare genetic disorders (sponsor NEGEDIA)
The field of genetic diseases has included numerous investigations targeting functionality and therapeutics, yet their success rates remain uncertain. To date, carrier status assessment, a potential approach for most genetic diseases, has encountered limitations due to cultural, technical, and economic factors. We aim to develop innovative, affordable, and scalable bioinformatics solutions for genomic surveillance and prevention of genetic diseases.
PI: Prof. Davide Chiacchiarelli (This email address is being protected from spambots. You need JavaScript enabled to view it.)
Develop machine learning approaches for the stratification of solid tumors (sponsor NEGEDIA)
The field of oncology is rapidly evolving towards the use of multigenic testing. Nevertheless, the use of RNA as a diagnostic, prognostic, and predictive tool is far from being adopted in clinical settings. We aim, based on existing IP in NEGEDIA, to develop and improve AI-based pipelines that leverages gene expression to stratify patients in multiple tumoral types, stages and progression.
PI: Prof. Davide Chiacchiarelli (This email address is being protected from spambots. You need JavaScript enabled to view it.)
Multi-omics approaches to integrate the layers of life
The onset of multiple -omics approaches has revolutionized the view of molecular biology. Nevertheless, a clear integration of -omics data is still missing in clinical and research settings. We aim to develop a joint approach to simultaneously profile genomics, transcriptomics, proteomics, lipidomics, and metabolomics from a single biological specimen, with the final aim of tracking gene expression from the genome to the phenome.
PI: Prof. Davide Chiacchiarelli (This email address is being protected from spambots. You need JavaScript enabled to view it.)
Unveiling structure, function and role of glycoprocessing enzymes by combined structural and molecular biology, bioinformatic and computational strategies.
The present PhD project proposal will decipher the structure and reaction mechanism of glycans processing enzymes from different gut microbes by a combination of structural and computational biology conjugated with bioinformatic approaches. By integrating these various “dimensions”, the project aims to yield innovative insights into the intricate world of carbohydrates chemistry and gut microbes. Structures, activity, binding specificity of adhesion and processing enzymes (glycoforms, glycosyl transferase (GT) and glycoside hydrolase (GH) families; adhesins) will be resolved in molecular detail by structural biology (NMR, native Mass Spectrometry, X-ray, EM) and computational, biophysical and bioinformatics approaches. Further informations are available at Silipo_project.docx.
PI: Prof. Alba Silipo (This email address is being protected from spambots. You need JavaScript enabled to view it.)
Title: Transfer learning methodologies for large-scale metagenomic analysis of the human microbiome
Research project: The human microbiome refers to the collection of microorganisms that live in our body. Its composition varies greatly between individuals and can be influenced by multiple factors including genetics, age, diet, lifestyle, and environmental exposure. Imbalance in the microbiome's composition has been also associated with various health conditions. Research on the human microbiome is still evolving, and predictive modelling based on machine learning approaches has become an important topic to answer to biological problems of interest. However, models trained on a specific set of data do not generalize well to other data acquired from different cohorts or laboratories due to the difference in the statistical properties of these datasets. To alleviate this problem, the project will focus on developing and implementing transfer learning methodologies (e.g., domain adaptation) with the aim of aligning the
statistical distributions before transferring models across datasets. The developed methods willbe validated in different cohorts acquired in the laboratory, as well as retrieved from public repositories, and involving the characterization of the gut and oral human microbiome from
large-scale metagenomic datasets.
PI: Prof. Edoardo Pasolli (This email address is being protected from spambots. You need JavaScript enabled to view it.)
Functional genomics and Bioinformatics for early detection of cell fate.
The project aims to the implementation of bioinformatics methodologies for early diagnostics and monitoring tools based on biomolecular data, to favor possible applications in biotechnology, ecology, biomedicine or agrifood.
We aim:
• to set up bioinformatics platforms for the collection of omics data from functional genomics, in order to define gene expression atlases that could describe the cellular responses in distinct biological systems (animals, plants, microbes: from humans to microorganisms) in different conditions;
• to develop analytical methodologies based on suitable data mining approaches and deep learning that, given the nature of the data (high dimensionality and complexity), represent essential tools for comparatively characterizing the complex network of cellular responses
during growth, development, and exposure to varied stimuli, in diversified biological systems. Further informations are available at CQB_Phd_Project_Proposal_2024_chiusano.docx.
PI: Prof. Maria Luisa Chiusano (This email address is being protected from spambots. You need JavaScript enabled to view it.)
Integrative bioinformatics for Trans-omics networks representation and image analysis
The aim of this proposal is to define novel integrative bioinformatics tools to provide a global view of different biological data sources and to unravel control systems of the cells. These tools will support the representation and analysis of biological processes by leveraging trans-omics networks, deep learning techniques and imaging analysis. The collection of biological data serves as a means to catalog the elements of life, but to truly understand a system, it requires the integration of these data using mathematical and relational models. These models can mechanistically describe the relationships between the components of the system, including the control systems of the cells and imaging data. Through advanced analysis of these datasets, we have made significant progress in understanding biological regulations, uncovering both generic rules and exceptions to these rules. These generic rules are observed as trends in the data, but there are often variations that deviate from these patterns.
To overcome the limitations of comprehensiveness and information gaps in interactions across multiple omic layers, we propose a trans-omics approach. This approach involves constructing a network structure that represents the biochemical trans-omics network, capturing causality and the input-output relationships at a molecular level. By incorporating control systems of the cells and imaging data into this network, we can gain deeper insights into biological processes.
Deep learning techniques offer promising tools for integrating multi-omics datasets and conducting various analyses based on the trans-OWAS approach and trans-omics representations. These techniques enable the identification of complex patterns and relationships within the integrated data, allowing us to uncover novel insights into cellular control systems and imaging analysis. Further informations are available at CQB_Phd_Project_Proposal_2024_Rinaldi.pdf.
PI: Prof. Antonio M. Rinaldi (This email address is being protected from spambots. You need JavaScript enabled to view it.)
Exploring Chemoresistance-Related Pathways in Neuroblastoma through Single-Cell Transcriptomics
Single-cell transcriptomics is a powerful tool that enables researchers to analyze gene expression profiles at the individual cell level, offering insights into cellular heterogeneity and the specific pathways involved in chemoresistance. This project aims to uncover the molecular mechanisms responsible for chemoresistance by examining individual neuroblastoma cell subclones through single-cell analysis. By performing computational analyses of single-cell RNA sequencing data from matched tumor samples taken at diagnosis and after chemotherapy, we will address crucial aspects of therapy resistance.
PI: Prof. Mario Capasso (This email address is being protected from spambots. You need JavaScript enabled to view it.)
Machine Learning and Artificial Intelligence to Characterize Tumor Heterogeneity at Single-Cell Resolution
Single-cell analysis has become a key technology in cancer research, providing unparallelize accuracy to study the heterogeneity and complexity of tumors. The volume and intricacy of single-cell RNA sequencing (scRNA-seq) data necessitate advanced analytical methods. This research aims to utilize Machine Learning (ML) and Artificial Intelligence (AI) to enhance the characterization of tumor heterogeneity at single-cell resolution. By applying these technologies to scRNA-seq data from malignant and non-malignant cells, we expect to develop accurate and efficient methodologies for identifying cancer cell subtypes, predicting gene expression, and uncovering novel insights into tumor biology. The new lab member hired within the CQB program will gain expertise on how to analyze large-scale single-cell data and how to build models that can be used to understand the effects of perturbations. The candidate will have access to state of the art computational infrastructures including high performance computing and multiple GPU. The research project will use data generated in house and data collected from public repositories. He will be integrated in the lab collaborating with multiple member in addition to the PI, including senior postdocs and phd students. The research activity will be will significantly improve the accuracy and efficiency of single-cell analysis in cancer research.
Expected outcomes include:
● Enhanced identification of cancer cell subtypes
● Improved gene expression prediction
● Discovery of novel cellular subtypes and gene expression patterns within tumors
● Insights into the tumor microenvironment and cancer heterogeneity
Publications on high-impact journal and participation to international conference is expected.
P.I.: Prof. Michele Ceccarelli (This email address is being protected from spambots. You need JavaScript enabled to view it.)
Biomaterial Design via Machine Learning Methods
The environmental challenges that we currently face require urgent definition of new sustainable biomaterials. In this context, amyloid fibrils generated as a result of the aggregation of proteins and peptides provide exquisite nanoscaffolds for a variety of applications. Traditionally these fibrils are however associated with neurodegenerative diseases, making it difficult to exploit them in biomaterial applications. It is therefore critical to define de novo peptide sequences able to form non-pathological amyloids to boost the next generation of biomaterials for sustainable biotechnologies. Further informations are available at CQB_De_Simone_Lombardi_XLdocx.docx
PIs: Prof. Alfonso De Simone (This email address is being protected from spambots. You need JavaScript enabled to view it.) and Prof. Angela Lombardi (This email address is being protected from spambots. You need JavaScript enabled to view it.)
Artificial Intelligence and Bioinformatics approaches for identifying Digital Biomarkers of oral cavity pathologies and for oral health prevention.
Poor oral hygiene increases the risk of contracting infectious diseases, from tooth decay to gum inflammation, to more serious chronic-degenerative conditions such as periodontitis, up to the most complex oral cancers that can affect different intraoral tissues. In this context, the integration of clinical, biological and physiopathological data, with the outputs of imaging methodologies and artificial intelligence models, is transforming the approach to prevention, diagnosis and treatment support of oral diseases, making it increasingly transversal and multimodal. However, the detection of intraoral diseases and anomalies is still largely based on visual examination supported by scanning techniques and radiographic imaging of the characteristics of the tissue and the lesion, exhibiting however a low sensitivity and reproducibility that do not allow for early diagnosis. The PhD project is part of this scenario, aiming at designing and developing a technological platform for the early diagnosis of oral cavity diseases through a multimodal approach and exploiting techniques, methods and tools of artificial intelligence (machine learning and deep learning) and bioinformatics. In particular, the development of data-driven models, starting from heterogeneous sources of information (bioimages, clinical history, chemical-biological data, etc.), that are explainable and interpretable would allow not only to explore the potential of such approaches in prevention and intraoral diagnosis but also to contribute to their translation into clinical practice.
PI: Prof. Mario Sansone (This email address is being protected from spambots. You need JavaScript enabled to view it.)
Enhancing Active Brain-Computer Interface Performance Using Neural Networks and Advanced Machine Learning Techniques
The direct interface between the brain and machines holds immense potential for advancements in healthcare and industry. However, the decoding of brain signals presents significant challenges, particularly in pattern recognition and real-time operation. While companies such as Neuralink invest heavily in invasive neuroimaging techniques, non- invasive methods like EEG remain more viable for widespread, daily-life applications. Consequently, advanced signal processing is crucial for accurately decoding brain-related time series data.
This project aims to explore the application of neural networks within a framework characterized by a low signal-to-noise ratio and limited data samples. We will investigate how neural networks, augmented by advanced machine learning techniques such as few-shot learning, meta- learning, and self-supervised learning, can enhance the decoding performance of EEG signals in these challenging conditions. These cutting-edge techniques are designed to optimize data utilization, manage inter-subject and intra-subject non-stationarity, and reduce the overall training costs and data dependency.
Our ultimate goal is to develop a neural interface that leverages spontaneous brain activity for applications in rehabilitation, industrial settings, and gaming. By integrating these advanced neural network techniques and modern machine learning approaches, we aim to overcome current limitations and push the boundaries of what is possible with non-invasive brain-computer interfaces.
PI: Prof. Francesco Isgrò (This email address is being protected from spambots. You need JavaScript enabled to view it.)