Information transduction capacity of mammalian cells is high and varies substantially from cell to cell.

The distributions of cellular neighborhood volumes in two very different multicellular species - snowflake yeast and Volvox carteri - are found to obey a common functional form arising from maximum entropy consideration, despite great differences in their cell division processes.

Emergent property inference, a novel machine learning methodology, learns distributions of neural circuit model parameters that produce computational properties and provides novel scientific insight through the quantification of the rich parametric structure it captures.

Unbiased and automatic annotation using structured prediction framework with efficiently built data-driven atlases is more accurate than registration-based methods for cell identifications in dense images and enables fast whole-brain analysis.

A data-driven network model offers an interpretable and physiologically sound description of the whole-brain spontaneous neural activity of zebrafish larvae.

Neurons differ in their impact on collective cortical activity, with sensitive neurons forming a stable topological core, implicated in cortical-state transitions, while peripheral insensitive neurons are more responsive to stimuli.

Combining powerful simulation methods uncovers the structural and dynamical changes driving G protein activation in atomic detail, revealing the allosteric network that triggers GDP release and reconciling diverse experimental data.

Evidence for neuromodulatory control of flexibility in human neural models.

Humans use a near-optimal eye movement strategy to efficiently extract information about high-level visual categories.

Sequence signatures can be used to discriminate between selected and non-selected immune repertoires at different stages only at the collective population level but not at the level of single cells.