Obesity is a worldwide epidemic, with major health and economic costs. Here we estimate heritability for body mass index (BMI) in 172,000 sibling pairs and 150,832 unrelated individuals and explore the contribution of genotype-covariate interaction effects at common SNP loci. We find evidence for genotype-age interaction (likelihood ratio test (LRT) = 73.58, degrees of freedom (df) = 1, P = 4.83 × 10), which contributed 8.1% (1.4% s.e.) to BMI variation. Across eight self-reported lifestyle factors, including diet and exercise, we find genotype-environment interaction only for smoking behavior (LRT = 19.70, P = 5.03 × 10 and LRT = 30.80, P = 1.42 × 10), which contributed 4.0% (0.8% s.e.) to BMI variation. Bayesian association analysis suggests that BMI is highly polygenic, with 75% of the SNP heritability attributable to loci that each explain <0.01% of the phenotypic variance. Our findings imply that substantially larger sample sizes across ages and lifestyles are required to understand the full genetic architecture of BMI.

Spontaneous activity is a fundamental characteristic of the developing nervous system. Intriguingly, it often takes the form of multiple structured assemblies of neurons. Such assemblies can form even in the absence of afferent input, for instance in the zebrafish optic tectum after bilateral enucleation early in life. While the development of neural assemblies based on structured afferent input has been theoretically well-studied, it is less clear how they could arise in systems without afferent input. Here we show that a recurrent network of binary threshold neurons with initially random weights can form neural assemblies based on a simple Hebbian learning rule. Over development the network becomes increasingly modular while being driven by initially unstructured spontaneous activity, leading to the emergence of neural assemblies. Surprisingly, the set of neurons making up each assembly then continues to evolve, despite the number of assemblies remaining roughly constant. In the mature network assembly activity builds over several timesteps before the activation of the full assembly, as recently observed in calcium-imaging experiments. Our results show that Hebbian learning is sufficient to explain the emergence of highly structured patterns of neural activity in the absence of structured input.

Discovering how neural computations are implemented in the cortex at the level of monosynaptic connectivity requires probing for the existence of synapses from possibly thousands of presynaptic candidate neurons. Two-photon optogenetics has been shown to be a promising technology for mapping such monosynaptic connections via serial stimulation of neurons with single-cell resolution. However, this approach is limited in its ability to uncover connectivity at large scales because stimulating neurons one-by-one requires prohibitively long experiments. Here we developed novel computational tools that, when combined, enable learning of monosynaptic connectivity from high-speed holographic neural ensemble stimulation. First, we developed a model-based compressed sensing algorithm that identifies connections from postsynaptic responses evoked by stimulation of many neurons at once, considerably increasing the rate at which the existence and strength of synapses are screened. Second, we developed a deep learning method that isolates the postsynaptic response evoked by each stimulus, allowing stimulation to rapidly switch between ensembles without waiting for the postsynaptic response to return to baseline. Together, our system increases the throughput of monosynaptic connectivity mapping by an order of magnitude over existing approaches, enabling the acquisition of connectivity maps at speeds needed to discover the synaptic circuitry implementing neural computations.

The circuit mechanisms that give rise to direction selectivity in the retina have been studied extensively but how direction selectivity is established in retinorecipient areas of the brain is less well understood. Using functional imaging in larval zebrafish we examine how the direction of motion is encoded by populations of neurons at three layers of the optic tectum; retinal ganglion cell axons (RGCs), a layer of superficial inhibitory interneurons (SINs), and periventricular neurons (PVNs), which constitute the majority of neurons in the tectum. We show that the representation of motion direction is transformed at each layer. At the level of RGCs and SINs the direction of motion is encoded by three direction-selective (DS) subtypes tuned to upward, downward, and caudal-to-rostral motion. However, the tuning of SINs is significantly narrower and this leads to a conspicuous gap in the representation of motion in the rostral-to-caudal direction at the level of SINs. Consistent with previous findings we demonstrate that, at the level of PVNs the direction of motion is encoded by four DS cell types which include an additional DS PVN cell type tuned to rostral-to-caudal motion. Strikingly, the tuning profile of this emergent cell type overlaps with the gap in the representation of rostral-to-caudal motion at the level of SINs. Using our functional imaging data we constructed a simple computational model that demonstrates how the emergent population of PVNs is generated by the interactions of cells at each layer of the tectal network. The model predicts that PVNs tuned to rostral-to-caudal motion can be generated via convergence of DS RGCs tuned to upward and downward motion and feedforward tuned inhibition via SINs which suppresses responses to non-preferred directions. Thus, by reshaping directional tuning that is inherited from the retina inhibitory inputs from SINs can generate a novel subtype of DS PVN and in so doing enhance the encoding of directional stimuli.