Search for: All records

Exciting advances in technologies to measure biological systems are currently at the forefront of research. The ability to gather data along an increasing number of omic dimensions has created a need for tools to analyze all of this information together, rather than siloing each technology into separate analysis pipelines. To advance this goal, we introduce a framework called the Single-Cell Multi-Modal GAN (scMMGAN) that integrates data from multiple modalities into a unified representation in the ambient data space for downstream analysis using a combination of adversarial learning and data geometry techniques. The framework’s key improvement is an additional diffusion geometry loss with a new kernel that constrains the otherwise over-parameterized GAN network. We demonstrate scMMGAN’s ability to produce more meaningful alignments than alternative methods on a wide variety of data modalities, and that its output can be used to draw conclusions from real-world biological experimental data. We highlight data from an experiment studying the development of triple negative breast cancer, where we show how scMMGAN can be used to identify novel gene associations and we demonstrate that cell clusters identified only on the scRNAseq data occur in localized spatial patterns that reveal insights on the spatial transcriptomic images.

The expansion of fluorescence bioimaging toward more complex systems and geometries requires analytical tools capable of spanning widely varying timescales and length scales, cleanly separating multiple fluorescent labels and distinguishing these labels from background autofluorescence. Here we meet these challenging objectives for multispectral fluorescence microscopy, combining hyperspectral phasors and linear unmixing to create Hybrid Unmixing (HyU). HyU is efficient and robust, capable of quantitative signal separation even at low illumination levels. In dynamic imaging of developing zebrafish embryos and in mouse tissue, HyU was able to cleanly and efficiently unmix multiple fluorescent labels, even in demanding volumetric timelapse imaging settings. HyU permits high dynamic range imaging, allowing simultaneous imaging of bright exogenous labels and dim endogenous labels. This enables coincident studies of tagged components, cellular behaviors and cellular metabolism within the same specimen, providing more accurate insights into the orchestrated complexity of biological systems.

The lateral hypothalamic area (LHA) integrates homeostatic processes and reward-motivated behaviors. Here we show that LHA neurons that produce melanin-concentrating hormone (MCH) are dynamically responsive to both food-directed appetitive and consummatory processes in male rats. Specifically, results reveal that MCH neuron Ca²⁺activity increases in response to both discrete and contextual food-predictive cues and is correlated with food-motivated responses. MCH neuron activity also increases during eating, and this response is highly predictive of caloric consumption and declines throughout a meal, thus supporting a role for MCH neurons in the positive feedback consummatory process known as appetition. These physiological MCH neural responses are functionally relevant as chemogenetic MCH neuron activation promotes appetitive behavioral responses to food-predictive cues and increases meal size. Finally, MCH neuron activation enhances preference for a noncaloric flavor paired with intragastric glucose. Collectively, these data identify a hypothalamic neural population that orchestrates both food-motivated appetitive and intake-promoting consummatory processes.

Malicious software (malware) classification offers a unique challenge for continual learning (CL) regimes due to the volume of new samples received on a daily basis and the evolution of malware to exploit new vulnerabilities. On a typical day, antivirus vendors receive hundreds of thousands of unique pieces of software, both malicious and benign, and over the course of the lifetime of a malware classifier, more than a billion samples can easily accumulate. Given the scale of the problem, sequential training using continual learning techniques could provide substantial benefits in reducing training and storage overhead. To date, however, there has been no exploration of CL applied to malware classification tasks. In this paper, we study 11 CL techniques applied to three malware tasks covering common incremental learning scenarios, including task, class, and domain incremental learning (IL). Specifically, using two realistic, large-scale malware datasets, we evaluate the performance of the CL methods on both binary malware classification (Domain-IL) and multi-class malware family classification (Task-IL and Class-IL) tasks. To our surprise, continual learning methods significantly underperformed naive Joint replay of the training data in nearly all settings – in some cases reducing accuracy by more than 70 percentage points. A simple approach ofmore »selectively replaying 20% of the stored data achieves better performance, with 50% of the training time compared to Joint replay. Finally, we discuss potential reasons for the unexpectedly poor performance of the CL techniques, with the hope that it spurs further research on developing techniques that are more effective in the malware classification domain.« less