Unimore logo AImageLab
 Vittorio Pipoli

Vittorio Pipoli

Position at AImageLab:

Research Fellow
Dipartimento d'Ingegneria "Enzo Ferrari", Modena Italy

Email:

vittorio_DOT_pipoli_AT_unimore_DOT_it

Phone:

+39 327 793 2983

Vittorio Pipoli

WORK EXPERIENCE
June 2022 – ongoing Research Assistant
University of Modena and Reggio Emilia, Enzo Ferrari Engineering Department (DIEF), Modena
(MO)
Collaborating on the Decider project (research project funded by the European Union from
1.2.2021 to 31.1.2026 under Horizon 2020 research and innovation programme under grant
agreement No 965193.) at Work Package 2 and Work Package 9.
– WP2: Developer of transformer-based architecture for multimodal learning from WSI and
omic data.
– WP9: Front-end developer exploiting Lit framework for the implementation of a tumor board.
EDUCATION AND TRAINING
March 2020 – April 2022 Master’s Degree in Data Science and Engineering (associated to "laurea
in Ingegneria Informatica (LM-32)")
Politecnico di Torino, Torino (TO)
Grade: 110/110
October 2016 – March 2020 Bachelor’s Degree in Computer Engineering
Politecnico di Torino, Torino (TO)
Grade: 95/110
September 2011 – July 2016 High school leaving qualification in scientific studies
Liceo Scientifico Leonardo da Vinci, Fasano (BR)
Grade: 100/100 with Honours

 

PUBLICATIONS
July 2022 V. Pipoli, M. Cappelli, A. Palladini, C. Peluso, M. Lovino, E. Ficarra,
Predicting gene expression levels from DNA sequences and posttranscriptional
information with transformers, Computer Methods and
Programs in Biomedicine
– This publication is a consequence of the Bioinformatics team project in collaboration with
Prof. Elisa Ficarra and Ph.D. Marta Lovino in the DAUIN department at Politecnico di Torino
(TO).
– In this team project, we developed several Deep Learning solutions to tackle a complex
regression task, predicting the gene expression levels by analyzing the aminoacid sequences
upstream and downstream of the transcription start site of the genes.
– I personally devised the transformer based architectures called TransformerDeepLncLoc,
which have overtaken our reference, which was Xpresso.
– The work has been accepted from Computer Methods and Programs in Biomedicine, and it
is currently in press.
– GitHub repository: https://github.com/geneexpressionpolito/Predicting-gene-expression-levels-from-DNA-sequences-and-post-transcriptional-info-with-transformers

-- DOI: https://doi.org/10.1016/j.cmpb.2022.107035

 

RESEARCH PROJECTS
November 2021 – April 2022 Master’s Thesis: Transformer-based architectures for long biological
sequences
Politecnico di Torino, Torino (TO)
– Transformer-based architectures for long biological sequences is a Master’s Thesis work that
deals with the employment of transformer-based architectures in gene expression prediction.
– In particular, I developed two different transformer-based architectures named ConvTransformer
and FNetCompression. The first reaches the best performances I can afford with
my means. In contrast, the second attempts to reach almost the same performance as the
former but compresses the information for more than 90%, leading to a huge complexity
reduction.
– My models outperformed state-of-the-art architectures such as Xpresso (on their dataset)
and can keep the comparison with ExPecto (on their dataset).

 


Research Activities


Publications

1 Pipoli, Vittorio; Cappelli, Mattia; Palladini, Alessandro; Peluso, Carlo; Lovino, Marta; Ficarra, Elisa "Predicting gene expression levels from DNA sequences and post-transcriptional information with transformers" COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE, vol. 225, pp. 107035 -107044 , 2022 | DOI: 10.1016/j.cmpb.2022.107035 Journal