Try tools, algorithms & atlases
developed in the Qutub Lab
Molecular & Cell Atlases
Leukemia Protein Atlases
2019 Nature Biomedical Engineering (main leukemia atlas); Molecular Cancer Research, 2018 (pediatric leukemias)
Purpose: Classification of leukemia patients (adult and pediatric) by protein signatures to help improve treatment and identify common signaling pathways in the hematological cancer
Benefits: Searchable web-based atlases for cancer research and potential therapeutic targets that maps patient outcome to specific protein signatures
Current applications: Discovery of new drug targets. Optimal matching of current therapies to best patient group. Fundamental insight into quantifiable cancer hallmarks.
Availability:
Atlases and corresponding de-identified clinical data are
available online at the following sites:
Acute Myeloid Leukemia Protein Atlas - Adult
Acute Myeloid Leukemia Protein Atlas - Pediatric
Acute Lymphoblastic Leukemia Protein Atlas - Pediatric
Main Page: https://www.leukemiaatlas.org/
MetaGalaxy Code: https://www.leukemiaatlas.org/code
Image Analysis
cytoNet
Purpose: Characterizes multicellular topology from microscopy images. Accessible via Amazon cloud, cytoNet takes as input color or binary images of cells or tissues, and quantifies the spatial relationships in cell communities using principles of graph theory.
Benefits: cytoNet identifies effects of cell-cell interactions on individual cell phenotypes.
Current applications: Understanding cell cycle dynamics in developing neural stem cells, characterizing the response of endothelial cells to neurotrophic factors present in the brain after injury
Availability:
GAIN Neuron Counting App
Purpose: Counting neural progenitor cells and neurons, and measuring dendrite outgrowth from individual cell bodies in images
Benefits: Overcomes limitations of applying prior algorithms to single cell analysis by a rule-based tracking method that maps soma to their dendrites. Interactive friendly graphical user-interface is integrated with the code. For use by all researchers.
Current applications: Automatically processing neuronal cell microscope images to help determine differentiation state of cells in different mechanical environments and in response to drugs.
Availability:
Interactive Data Visualization
Biowheel
bioRxiv 2018
Purpose: Visually interpret high-dimensional data through interactive graphs
Benefits: Ease of use & speed. No programming needed. Drag & drop files into the cloud-based tool. Collaborate on biomedical data science projects.
Current applications: Collaboration on data challenges. Expert-informed learning. Teaching tool for clustering and pattern recognition. Clinical decision-making.
Clustering & Cluster Optimization
Shrinkage Clustering
BMC Systems Biology, 2018; Leibniz-Zentrum für Informatik Proceedings, 2017
Purpose: Identifying the ideal number of groups (e.g., cancer patients, cells, social networking groups).
Benefits: Very fast, computationally efficient method that enables minimal cluster size constraints.
Current applications: Classifying cells and tissues. Designing clinical trials. Applications to TCGA, Allen Brain Institute, and Wisconsin breast cancer datasets illustrated in the BMC Systems Biology paper.
Availability:
Shrinkage Clustering is available open source as an R package on Github. Credits: Quentin (Hanyang) Li, Wendy Chenyue Hu
Progeny Clustering
Scientific Reports, 2015; The R Journal, 2016
Purpose: Identifying the ideal number of groups (e.g., cancer patients, cells, social networking groups).
Benefits: Computationally efficient. Conserves on the number of samples needed. Compatible with multiple clustering approaches.
Current applications: Designing clinical trials. Identifying the effectiveness of human cell patterning.
Availability:
Progeny Clustering code package is available open source in the R repository: “progenyClust: Finding the Optimal Cluster Number Using Progeny Clustering”, as described in (Hu & Qutub, R Journal 2016). A user-friendly outline of the progenyClust R package is available online here thanks to Ian Howson.
Metabolic Modeling
Whole tissue metabolic models
CORDA2 library
Pacific Symposium on Biocomputing, 2017
Purpose: Provides a method to produce cell and tissue-specific metabolic models
Benefits: Predicts metabolism in healthy and diseased cells based on experimentally-obtained molecular expression data. Faster than CORDA and noise-independent. Optimized for comparing subtypes of human tissues and cells.
Current applications: Identifying differences in metabolism within subtypes of cancer and healthy mammalian cells. Generating patient-specific models.
Availability:
Library (in MATLAB format)
CORDA2 function file as provided in the paper’s supplemental information
Matrix-Form Artificially Centered Hit and Run, mfACHR
Pacific Symposium on Biocomputing, 2017
Purpose: Provides a method to improve upon Monte Carlo Sampling of high dimensional network models
Benefits: Speeds up sampling of a model’s solution space compared to prior algorithms.
Current applications: Identifying steady-state metabolic flux distributions within cell and patient-specific whole genome wide metabolic models.
Availability:
mfACHR function file (in Matlab format) as provided in the paper’s supplemental information.
CORDA library
Purpose: Provides a library of tissue specific metabolic models
Benefits: Predicts metabolism in healthy and diseased tissues based on experimentally-obtained molecular expression data
Current applications: Identifying differences in metabolism across cancerous and healthy mammalian tissue
Library availability:
CORDA function file (in Matlab format) as provided in the paper’s supplemental information
Python version of CORDA developed by Christian Diener
corsoFBA
Purpose: Models the flux of metabolites through tissue.
Benefits: Overcomes biomass production assumptions of other FBA methods.
Current applications: Modeling the metabolism of mammalian cells, and changes in disease.
Availability:
corsoFBA Matlab m-file (description)
Signaling Models
Protein signaling pathway models
HIF1 hydroxylation chemical-kinetic models
J Cell Sci, 2006; BBA Mol Cell Bio, 2007; Mol Cell Biol, 2008
Purpose: Models post-translational regulation of hypoxia-inducible factor 1 (HIF1), a protein activating hundreds of genes as a function of oxygen availability
Benefits: Predicts levels of HIF1 as a function of its cofactors
Current applications: Helping optimize the design of experimental modulation of hypoxia-inducible factor proteins in cancer and neural progenitor cells
Availability:
Data Challenges
Acute Myeloid Leukemia Outcome Prediction Challenge (DREAM 9)
PLOS Comp Bio, 2016
DREAM 9 Wiki site, 2014-2015
Breast Cancer Challenge (DREAM 8 HPN)
Nature Methods, 2016
DREAM 8 Wiki site, 2013-2014
Breast Cancer Proteomics Visualization Portal
