DEVIL accelerates large-scale genomic analysis by up to fortyfold
More than ten million cells analysed in less than two hours, with memory requirements approximately three times lower than those of the best existing tools and processing speeds up to forty times faster on the largest datasets. This remarkable achievement was made possible by a team of researchers from the University of Trieste, Area Science Park, SISSA and Human Technopole, who developed DEVIL (Differential Expression with Variational Inference Learning), a new high-performance computational tool. The study has been published in Nature Communications.
Understanding which genes are active within cells is fundamental to advancing our knowledge of disease and developing new therapeutic approaches. Today’s most advanced technologies enable researchers to measure gene activity in millions of cells from dozens or even hundreds of patients, generating unprecedented volumes of data for biomedical research. However, this revolution also brings with it two major challenges: on the one hand, the risk of errors in data interpretation, and on the other, the difficulty of analysing such large volumes of information.
The first challenge is computational. Analysing millions of cells requires enormous computing power. Traditional methods are too slow and consume excessive memory when handling datasets of this scale, creating a bottleneck that risks undermining the advantages offered by modern data collection technologies. The second challenge is statistical. Cells collected from the same patient tend to resemble each other more closely than cells from different patients, as they share the same biological background, environment and individual characteristics. Ignoring this aspect—as many currently available tools do—can lead to biased statistical conclusions, with the risk of identifying cellular changes as “significant” when they are not, or conversely failing to detect genuine ones.
To address both issues, the researchers developed DEVIL, successfully combining statistical rigour with computational speed in an innovative way. From a computational perspective, DEVIL, which was also developed with support from the AIRC Foundation, has been specifically designed to efficiently exploit the latest parallel computing architectures commonly used in artificial intelligence. Moreover, DEVIL is not only faster but also significantly more memory-efficient. This is a crucial advantage, as analyses previously restricted to major computing centres can now be carried out by smaller research laboratories and infrastructures. From a statistical perspective, DEVIL tackles the problem through a Bayesian approach that correctly accounts for the structure of the data, treating cells from the same patient as correlated observations and thereby distinguishing differences between patients from genuine differences in cellular activity.
“This work would not have been possible without ORFEO, Area Science Park’s high-performance computing and data centre, recently upgraded through funding from Italy’s National Recovery and Resilience Plan (PNRR),” said Stefano Cozzini, Director of the Institute for Research and Technological Innovation at Area Science Park. “The availability of state-of-the-art GPUs, offering exceptional computing performance, together with the optimisation of algorithms carried out by our team for this architecture, has enabled DEVIL to tackle problems on a significantly larger scale than previously possible. This achievement is the result of a highly skilled multidisciplinary team that has been able to fully leverage the capabilities of the infrastructure and technologies available”.
”Differential expression analysis, —that is, the statistical method used to identify which genes are significantly more or less active across different biological conditions—is a mature technology,” explained Giulio Caravagna of the University of Trieste. “However, the transition to single-cell data has introduced statistical and computational challenges that make the integrated analysis of large patient cohorts particularly difficult. Our work was conceived specifically to overcome this bottleneck, combining methodological innovation with high-performance computing to enable the analysis of millions of cells from hundreds of patients”.
“In the development of DEVIL, the synergy between classical and Bayesian statistical methods represents a distinctive strength within the oncology literature,” added Leonardo Egidi of the University of Trieste. “It makes DEVIL both a highly efficient computational protocol and a methodologically robust one. Future developments may involve spatio-temporal models for multiple patients and further computational approximations based on theoretical properties that are currently under investigation—a fascinating combination of statistical, computational and biological expertise”.
DEVIL was validated on two real-world biological case studies. In the first, focused on the identification of immune cell populations, the tool demonstrated greater accuracy and specificity in detecting biologically relevant functions. In the second, which investigated the ageing of human muscle tissue, DEVIL identified age-related transcriptional changes in a more robust and biologically meaningful manner, reducing noise and highlighting key processes for downstream analyses.
Released as free and open-source software, DEVIL is now available to laboratories and hospitals worldwide, paving the way for a new generation of large-scale genomic analyses to support cancer research, the study of degenerative diseases and the advancement of personalised medicine.