Summer
Undergraduate
Research
Experience in
Bioimaging
Engineering &
Technology

Home

Graduate Training Opportunities

Mission

Faculty

Students and Postdocs

Advisory Board

Publications

Presentations

Image Collections

SUREBET summer program

News and Awards

Bioimaging Day

CBI at UCSB

Map

Jelena Kovacevic and Robert F. Murphy, Directors

2006 SUREBET Participants

Imaging of Mechanically Induced Calcium Mobilization
Calcium performs a distinctive role in cell function. The purpose of this project was to measure the relative amount of calcium in cells subsequent to chemical and mechanical stimulation. The addition of these stimuli caused calcium to be released into the cell's cytosol. Following release, cells returned their basal calcium level. For the purpose of detecting calcium in NIH 3T3 cells, conventional epi-fluorescent microscopy was introduced. The fibroblast cells were loaded for 15 minutes with 2µM fluorescent calcium dey, Fluo 4-AM. The dye enters the cell freely due to the lipophilic AM group and is trapped intracellularly when endogenous esterases cleave the group from the dye molecule. Live cell images were taken every 5 seconds to detect any response to stimuli. As a chemical agonis, 100 µM ATP, which stimulates the cell's purinergic receptors, was added to the cell's environment. Many cells responded to this chemical stimulation; this response is visualized as a brighter glow in cell images and in analytical data, a spike. Although there was possible interference from sources such as noise and photobleaching, robust calcium responses were observed. For the mechanical stimulation, elementary magnetic tweezers were constructed consisting of a magnetized needle and ferromagnetic particles coated in fibronectin and bonded to the cell's integrin receptors. However, due to technical difficulties with the apparatus, the only result achieved was an actual movie of a cell being ripped off the substrate. However, the cells did respond to this mechanical stimulation with an increase in dye brightness, although the purpose of gently tugging on the cells to detect a calcium response was defeated. Future work for this project will include the perfection of the mechanical force; the introduction of gadolinium, a stretch-activated calcium channel inhibitor; and application of ionomycin which chaperons calcium outside the cell into the cell.



Ashley Gonzalez-Lopez
Universidad Metropolitana
Sophmore, Cell and Molecular Biology
Graduate Student Advisors: Shann-Ching Chen and Elvira Garcia Osuna
Faculty Advisor: Dr. Robert F. Murphy

Image Acquisition for Heterogeneous Protein Location Pattern Analysis
The subcellular location of proteins is critical to the understanding of cell function. Comprehensive analysis is needed as part of systems biology efforts to understand the behavior of all expressed proteins. This location can be determined by the interpretation of fluorescent microscope images, and automated systems to this have been described. However, these have used primarily single cell images and do not take neighboring cell information into account. This is valuable information that will increase the accuracy of these systems (Chen and Murphy, 2006). Therefore, our goal is to improve classification accuracy by obtaining multiple cell images with two location patterns. The cells that were used were NIH3T3 cells, which have been GFP-tagged. We chose two protein patterns and tagged one of them with a membrane dye (Vybrant CM-Dil). The cells of both patterns were then grown together. We used the membrane dye as ground truth in the computational analysis. We have found an appropriate concentration of the dye. After two days of growth, Hocchst33342 (DNA dye) was added to the cells at an appropriate concentration. The cells were then imaged using an automated fluorescent microscope. We determined that the dye does not transfer between different cells. Therefore we can conclude that the dye is a good marker for determining the ground truth. Considering multiple cell images with heterogeneous protein location pattern may be expected to improve discrimination between similar patterns.



By: Jennifa Mohammed
University of the Virgin Islands
Senior, Computer Science
Graduate Student Advisor: Charles Jackson
Faculty Advisor: Dr. Jelena Kovacevic

Efficient Acquisition of Fluorescence Microscopy Images
We present proposed algorithms for the efficient acquisition of fluorescence microscopy datasets. We evaluate these with a protein classification system based on subcellular location patterns. The goal of this project is to acquire the datasets more efficiently in terms of the speed at whcih we acquire images, as well as the amount of light that the specimen is exposed to, in order to speed up acquisition and reduce photobleaching. We compared downsampling, reduced resolution, and reduced exposure time. our results found that downsampling is the most efficient of these methods, but that a combination of all three can yield even better results.



Julian Ortiz-Perez
Universidad Metropolitana
Senior, Computer Science
Graduate Student Advisor: Zhen Zhen Kou
Faculty Advisor: Dr. William Cohen
Classifying Captions Based on the Model Organism for SLIF

Subcellular Location Image Finder (SLIF) is a system which extracts information from both images and the associated captions in biological journal articles. This system SLIF interprets fluorescence microscope images containing the patterns of proteins within cells, and associates those patterns with the names of proteins and cell types in the accompanying caption. SLIF can generate assertions such as "Figure N depicts a localization of type L for protein P in cell type C". The goal of SLIF is to develp a large library of annotated and analyzed fluorescence microscope images, in order to support data-mining. The goal of this project was to build a new module for SLIF that will take a figure caption and predict which "model organism" (experimental animal) is being studied by the biologist in the paper. To start with, we will build three classifiers that predict: are mouse cells being studied (yes or no)? Are drosophila (fruitflies) being studied (yes or no)? Are yeast cells being studied (yes or no)? We will use machine-learning based text classification methods here. so the steps are: For each organism, collect positive and negative examples of papers about that organism. I plan to collect 100 positive and 100 negative examples to train each classifier. I will start with Naive Bayes algorithm. I plan to start with bag-of-word features and add some image features if necessary. For each organism, train a text classifier, and evaluate how accurate its predictions are. I will integrate the classifiers with the best accuracies into SLIF.



Jose Reyes
Universidad Metropolitana
Senior, Computer Science
Graduate Student Advisor: Justin Newberg
Faculty Advisors: Dr. Robert F. Murphy and Dr. Justin Crowley
Automated Recognition of Dendritic Spines in in vivo two-photon Laser Scanning Microscopy Datasets

Axonal boutons and dendritic spines play an important role in neural plasticity. They are the bridges between axons and dendrites, so changes in their distributions affect the connectivity between neurons in the brain; since the ability for the brain to learn new things or recover from trauma (such as stroke) are known to change connections in the brain, it is important to analyze boutons spines to better understand neural plasticity. historically, analysis of neural arboration has been challenging because of the vast number of dendrites and spines in a neural network, so there is a need for automated analysis of neuronal connectivity. In our project, we modify code developed by the Roysam laboratory which traces axons and dendrites in 3D fluorescence micrographs to detect spines and boutons. Analysis is performed on visual cortex image stacks from live mice, obtained with two photon laser-scanning microscopy. Labeling of spines and boutons will be compared to hand labeled traces in order to determine the effectiveness of tracing.