Unravelling the 3D-BrAIn data
Introducing Cristina Campi, University of Genova
Developing methods for data analysis may seem like a trivial task nowadays, at a time when artificial intelligence is ruling. However, expertise is needed to extract meaningful information out of the overwhelming amount of data that surround us. This is particularly true for complex experimental data measured with new technologies, such as the data collected from the 3D-BrAIn organoids. In 3D-BrAIn, Cristina Campi and Michele Piana from University of Genova (UNIGE) bring their mathematical modelling expertise to the table to decode and give meaning to the brain signals generated in the 3D-BrAIn lab.
UNIGE’s role within the project is to develop methods for the analysis of data collected from the brain organoid’s activity. This involves several steps. “Our first step is to assess the quality of the data, and clean it up in such a way to eliminate redundant and superfluous information,” Cristina Campi explains. “We first need to be sure about the quality of the data, to know whether we are looking at ‘real’ signals from the organoid or if we are looking at noise.” Moreover, thanks to using this advanced 3D multi-electrode array (MEA) to measure the organoid’s activity, these signals can be measured at an extremely high rate, which means the data resulting from these measurements is extremely large. With such a high amount of signals, it is easy to get lost in the meaning of every signal measured. Cristina: “We want to extract the meaningful features from this data, using different machine learning techniques.”
After the initial cleaning step, the next step is categorizing and labelling the data. To do this, we need to find similarities and differences between the signals, and cluster these signals together to label them. Labels for signal groups can for example be a healthy organoid, a diseased organoid, or an organoid receiving a certain treatment. The ultimate goal is that the machine-learning model can identify the correct label for the data itself, so-called ‘unsupervised machine learning’. “Before that, we are going to implement some supervised machine learning methods, where the artificial intelligence algorithms are trained using the labels. This makes sure that the model has a good capacity of distinguishing between the different labels, and hence a good performance in assigning one of the labels to new data that has not been labelled before,” Cristina explains.
“If we will be able to validate these models, we may use them to evaluate the effect of possible treatments, by assessing whether the signals measured in pathological organoids after a certain treatment are more similar to those of a healthy one.”
“In this way, we may be able to distinguish, for example, the signals measured by an organoid representing a certain pathology from those obtained from a healthy organoid. If we are able to validate these models, we may use them to evaluate the effect of possible treatments, by assessing whether the signals measured in pathological organoids after a certain treatment are more similar to those of a healthy one.”
With a background in Mathematics, Cristina Campi currently works as an Associate Professor of Numerical Analysis at UNIGE. She is also a member of the Method for Image and Data Analysis (MIDA) group and the Life Science Computational Lab (LISCOMP lab), a joint laboratory between UNIGE and the San Martino Polyclinic Hospital in Genova. Cristina Campi mentions: “As a professor at the Department of Mathematics at UNIGE, you never get bored! Our duties cover teaching activities, research and sometimes paperwork and institutional activities, like sitting on thesis defence committees. My research activity is carried out both at the Department of Mathematics, as well as at the San Martino Polyclinic Hospital, where the LISCOMP lab is located. There, I have the chance to meet with other researchers and physicians, and work on medical data where they are collected or measured.”
“… a key component for successful research work is the willingness to learn, and sometimes compromise to the scientific languages and backgrounds of the other researchers.”
Having worked in different research groups, Cristina highlights the importance of a multidisciplinary project: “During my formation years I have been extremely lucky to work in multidisciplinary research groups, where I understood that a key component for successful research work is the willingness to learn, and sometimes compromise to the scientific languages and backgrounds of the other researchers. I really look forward to collaborating with the other partners, in particular with the young researchers and PhD students that have been recruited so far and that will join us during the project.”
“Hopefully with long-term spin-offs in the clinical field as well, the 3D-BrAIn bio-digital brain twin can provide the community with a tool capable of testing new therapies in a controlled and safe environment.”
On the overall impact of the 3D-BrAIn project, Cristina highlights the technological and preclinical aspect, as well as the importance of the clinical aspect. “Hopefully with long-term spin-offs in the clinical field as well, this can provide the community with a tool capable of testing new therapies in a controlled and safe environment, such as the bio-digital twin that is to be studied in 3D-BrAIn.”
UNIGE’s prediction models will be a key part in the development of the 3D-BrAIn bio-digital brain twin, to offer a safe and controlled environment to test new therapies for neurological diseases. To achieve this stage, it first requires unravelling the data from the 3D-BrAIn model. UNIGE is working hard to create clarity in these large amounts of complex data.
Make sure to stay tuned and follow our journey towards this possible tool for testing new therapies.
