Biotech Innovations

Engineered biological systems, and the computational work that has to happen before anything is built.

A six stage map of where modelling enters CAR-T cell therapy, from antigen receptors and treatment specificity through combination therapy, time and dosage, cell dynamics and treatment efficacy, running from cancer to remission. — Where a model can carry weight in CAR-T cell therapy, stage by stage, from the receptor to the response. Each stage is a place where a question can be asked of a model before it is asked of a patient. Putignano G et al., Frontiers in Immunology 2025. CC BY.↗

What we work on

The question this group works on is which parts of a biological design can be settled in a model, and which still have to be settled at a bench.

That covers engineered cell therapies, where the behaviour of a construct can be reasoned about mathematically long before it exists; digital twin approaches, where a model stands in for a system that cannot be observed directly; and the multi-omic and clinical data that tell you whether the model resembled anything real.

Methods

No method is right for every question, and picking the wrong one wastes the study rather than the afternoon. The matrix below is our own published mapping of model families to the kinds of question each can answer.

Ordinary differential equations

Signalling pathways and cell population dynamics, where the mechanism matters more than the pattern, and where a parameter has a meaning you can argue about.

Agent-based models

Spatial interactions between a tumour and an immune system. Behaviour that emerges from many local rules rather than from one equation.

Machine learning

Complex patterns in data too large to read. Strong at prediction, weak at telling you why, which is why it is one method here and not the method.

Bayesian methods

Prediction that carries its own uncertainty, and population parameters estimated from small or uneven data. The method of choice when the honest answer is a range.

Control theory

Feedback systems, and asking what a therapy should do next given what it has done so far.

Sensitivity analysis

Which parameters actually move the result. Run first, because it tells you which of the others are worth the effort.

A matrix of six model families against four research question types: mechanistic understanding, outcome prediction, parameter estimation and system optimisation, each cell rated for how well that model family suits that kind of question. — Six model families against four kinds of research question. Reading a row tells you what a method is for. Reading a column tells you which method to reach for. Putignano G et al., Frontiers in Immunology 2025. CC BY.↗

Principal investigator

Guido Putignano

President and Founder

Computational biology, from engineered cell therapies to population studies on administrative health data.

3 publications, and what he works on →

People in this area

Alessia Soru

Scientific Projects Lead

PhD student in Oncology, Hematology and Pathology at the University of Bologna, and Board Member of Women&Tech® ETS.

Olufemi Olusola

Scientific Projects Lead

Biostatistician working at the intersection of agentic AI, clinical data, and digital twin therapeutics.

Roberto Russo

Student Ambassador

High school student, working with the Foundation's operational team.

Daniela Marotto

Advisory Board

Rheumatologist

Rheumatologist and leader in Italian health science, recognized for her commitment to multidisciplinary care.

Research and publications →

Working on something close to this?

We are looking for collaborators, not customers. If one of the questions above is one you are already working on, write to us.

Start a conversation →