Bajaj Lab (Chandrajit Bajaj, CS)
Bajaj's research group focuses on high throughput and big data analytics to elucidate, predict and validate molecular-molecular structural interactions for use in molecular therapeutics. Molecular sequence and structural imaging (1D, 2D) data come from experimental apparatus such as gene arrays, x-ray diffraction, and electron microscopy. These datasets are inherently noisy, and they must be computationally analyzed to construct 3D structural models, using highly regularized solutions to complicated inverse problems. The innovativeness of their approach is to always work with compressed data using and producing novel goal-directed image compression methods.
Dhillon Lab (Inderjit Dhillon, CS and Math)
Dhillon's research group focuses on developing novel solutions for Big Data Analytics that arise in various modern applications. These include models, algorithms and software that scale to very large data sets for various classical analysis tasks, such as regression, classification, dimensionality reduction as well as new analysis tasks, such as multi-task learning, matrix completion and high-dimensional covariance estimation. In particular, they apply these analysis tasks to social network analysis, recommender systems, bioinformatics and neuroscience. In a joint project with biologists, his group has developed new network-based methods that predict novel gene-disease associations based on a network of known associations between genes, human diseases and phenotypes of model species. In a joint project with neuroscientists, his group has developed scalable methods to analyze large-scale fMRI data by estimating a sparse inverse covariance matrix using new techniques from high-dimensional statistical inference. More generally, his research group has been developing software tools that are being used in various applications that involve big data analytics, such as non-negative matrix factorization, inverse covariance selection, high-dimensional clustering and co-clustering.
Press Lab (William Press, CS and IB)
Press's research group works on large-data problems in biology using a variety of signal-processing and statistical algorithms. Examples include the development of new preparation protocol and informatic pipeline that can achieve order of magnitude higher accuracy from next generation sequencing; studies of the comparative genomics of ultra-conserved noncoding DNA in mammals; comparative genomics of multiple chiroptera species that may be reservoirs of Ebola hemorrhagic fever; study of possible new protocols for adaptive clinical trials based on Bayesian bandit problems; and informatics pipelines for shotgun proteomics and mass spectrometry.