You can access the Lab's Github Repository by clicking the link below


Lab Projects

1. MAGIC (Markov Affinity-based Graph Imputation of Cells): an algorithm that uses graph signal processing for denoising and missing transcript recovery in single cell RNA sequencing. MAGIC successfully recovers gene-gene relationships from data and allows for the prediction of transcription targets.

2. PHATE (Potential of Heat-diffusion Affinity-based Transition Embedding): a visualization and dimensionality reduction technique that is sensitive to local and global relationships and successfully preserves structures of interest in biological data including cluster structure and branching trajectory structure.

3. SAUCIE (Sparse AutoEncoders for Clustering Imputation and Embedding): a deep autoencoder architecture that allows for unsupervised exploration of data, and has novel regularizations that allow for data denoising, batch normalization, clustering and visualization simultaneously  in various layers of the network.

4. DREMI (conditional Density-based Resampled Estimate of Mutual Information): a computational method based on established statistical concepts, to characterize signaling network relationships by quantifying the strengths of network edges and deriving signaling response functions.

5. DyMoN (Dynamics Modeling Networks): A neural network framework for learning stochastic dynamics for generative and embedding purposes. DyMoN serves as a deep model that itself embodies a dynamic system such that the gene logic and features driving the system can be studied.

6. SUGAR (Synthesis Using Geometrically Aligned Random-walks): A new kind of data generation algorithm that generates of data geometry rather than density and is able to generate data in sparse regions and predicts hypothetical data points.

7. MAGAN (Manifold Alignment Generative Adversarial Network): a  dual GAN (generative adversarial network) framework that can find correspondences between two data modalities measuring the same system to create an integrated dataset.

8. Neuron Editing: a neural network inference method which maps between datasets with the signal learned on a subset of the data, for example when a drug perturbation is applied to one cell type neuron editing can be used to apply it in silico to another cell type.