Salvatore Milite

🔗 This is a very small step, but I believe that these autodiff agent-based models provide a nice way of simulating biological ensembles with some level of physical constraints and realism. On that line, see also similar work with cell Potts models. arxiv.org/abs/2502.02129

📊 Compared to traditional agent-based modeling with ABC inference, we achieve better results with less manual tuning. In complex domains, coming up with a reasonable set of actions for the ABC is hard.

📄 In this work, we try to bridge two parallel research areas: the well-established literature on agent-based modeling in biological systems and recent advances in learning update rules in CA from data with autodifferentiation. arxiv.org/abs/2506.20486

How do you make sense of single cell multi-modal data beyond clustering? We need better methods that incorporate biological principles! Check out our latest method MIDAA, deep learning for single-cell multiomics with evolutionary principles: genomebiology.biomedcentral.com/articles/10.11…

[t.ly/vYqW7] Cancer biomarker discovery is historically mutation-centric, often neglecting the key role of genomics instability. Recently, we asked ourselves how do mutations and copy number alterations (CNAs) determine prognosis and tropism? 😋😋1/n