2015 Student Abstracts
- BREAKOUT SESSION I 8:30-8:50 a.m.
- BREAKOUT SESSION II 8:55-9:15 a.m.
- BREAKOUT SESSION III 9:20-9:40 a.m.
- BREAKOUT SESSION IV 10-10:20 a.m.
- BREAKOUT SESSION V 11:40 a.m.-12 p.m.
- BREAKOUT SESSION VI 3:15-3:35 p.m.
BREAKOUT SESSION I, 8:30-8:50 a.m.
Systematic Rational Identification of Sex-Linked Molecular Alterations and Therapies in Cancer
Jonathan Ma and Sadhika Malladi, Class of 2016
Research Conducted At: Harvard Medical School, Department of Pathology
Research Mentors: Dr. Andrew Beck
Harker faculty mentor: Mr. Christopher Spenner
2014 Siemens Regional Finalist
Though patient sex influences response to cancer treatments, time-consuming trial-and error screening currently hinders prediction of sex-disparate drug effects. We developed a novel framework combining genomic, pathway and connectivity map analyses to rationally predict sex-disparate molecular alterations and treatment responses. Through analyses of genomics data collected from thousands of patients, we identified significant genomic differences between the sexes in 17 cancers as well as between neoplastic and nonneoplastic tissue within each sex in seven cancers. Using these genomics results, we discovered sex-disparate pathway associations with neoplastic expression. Constructing sensitivity and resistance signatures from our genomics results, we then used connectivity maps to predict perturbagens to which both sexes are sensitive and perturbagens to which one sex is sensitive while the other is resistant in each of seven cancers. Notably, we correctly predicted that females are sensitive and males are resistant to tamoxifen treatment of lung adenocarcinoma. Furthermore, we predicted that tumors are equally sensitive in both sexes across all seven cancers considered to inhibition of cyclin-dependent kinase activity, for which sex-specific effects are untested. Thus, our approach is a valuable tool for large-scale discovery of sex-differentiated molecular alterations and for rational prediction of sex-disparate and sex-independent perturbagen sensitivity and resistance.
A Fusion Gene Detection Pipeline for the Identification of Gene Fusion Candidates in Cancer Cell Lines Using RNA-Seq Data
Vivek Sriram, Class of 2015
Research conducted at: Stanford University, Department of Biochemistry
Research Mentors: Dr. Julia Salzman, Mrs. Linda Szabo
Fusion genes, gene hybrids that form from translocations of nucleotide sequences in the genome, have been correlated with the presence of oncogenes. Fusion genes that drive solid tumors can offer remediable targets for precision therapies and can play significant roles as cancer biomarkers. A fusion gene detection pipeline was designed to provide a comprehensive methodology for discovering novel fusion genes. This consisted of a series of computational data analysis steps to read RNA-Sequencing data on cancer cell lines and predict potential fusion genes present. The pipeline is primarily written in the Python programming language and has been comprehensively tested on 64-bit Linux machines and on 16-bit MacOS platforms. The pipeline is built extensively around mapping and search algorithms and uses different data structures relevant to processing steps, including the memory-efficient interval tree data structure. To eliminate spurious fusion reads, a statistical method called False Discovery Rate was used to limit the presence of false positives. The performance of the pipeline was evaluated for efficiency, sensitivity and specificity. The pipeline filtered and identified hundreds of potential fusion genes in each of seven different publicly available cancer cell lines, with millions of records each, in a computationally efficient manner. This pipeline is unique in its ability to handle both RNA-Sequencing paired-end and single-end read data. Plugging such a comprehensive pipeline into the typical diagnosis cycle could be a further step in the identification of cancer genes, which will facilitate improved patient treatment and survival.
The PainGauge: A Feasibility Study Comparing the Effectiveness of a Mobile Health App and a Pain Diary in Assessing Pain
Divya Periyakoil, Class of 2016
Research Conducted At: Individual Mobile Health Research Project
Research Mentor: Mr. Ravi Pamnani, MS
Harker faculty mentor: Mr. Christopher Spenner
The PainGauge is a platform-agnostic, HIPAA secure, mobile-health application that can be accessed through any smartphone, iPad, laptop or desktop and used by patients with chronic pain to transmit pain scores to an encrypted, secure server. Data can be easily retrieved by providers or caregivers (in the form of a graph or table) using a password. The functionality, usability and user-friendliness the PainGauge App was tested with seven older adults to compare its effectiveness against a traditional pain diary. The PainGauge app was installed in each participant's smartphone. They were asked to record their pain scores at least once every two hours using the app or a diary. All participants utilized both modalities in random order for three days each. Data included their pain scores and user feedback at the end of the study period. Participants recorded more data with the PainGauge (98 data points with accurate time codes) than the diary (54 data points and many missing time codes). All participants preferred the PainGauge as it was easy to use and just required them to press a button and automatically recorded the date and time of the pain scores, minimizing their burden. One participant recommended that the PainGauge should also be available as a “small device worn around the neck with buttons to press to record pain scores" for patients who cannot/do not use smartphones." Older adults found PainGauge mHealth App easier to use and preferred it over the traditional pain diary.
Managing the Impact of Infrastructure Projects on Endangered Species: A Stochastic Simulation Approach Based on Population Viability Analysis
Venkat Sankar, Class of 2017
Research Conducted At: Harker Labs
Faculty Research Mentors: Mr. Daniel Sommer, Mr. Jeff Sutton
As the global population grows, it has become increasingly common for infrastructure projects such as dams, highways and solar farms to adversely impact the habitat quality and population viability of endangered species. Current field survey based approaches to environmental impact assessment lack rigor and detail, and these shortcomings could result in substantial environmental damage. The goal of this research is to develop a stochastic simulation based approach to systematically assess and manage this impact. Using a solar farm project in California and its impact on the endangered Giant Kangaroo Rat (Dipodomys ingens, GKR) as a real-world case study, we constructed and validated a simulation model based on Population Viability Analysis (PVA) as a rigorous and cost-effective approach. In the case study, we were able to establish that the solar farm as proposed will have a significant adverse impact on the GKR population, with an estimated >40 percent likelihood of GKR extinction over 100 years. Also identified were alternative layouts for the solar farm that could significantly reduce the extinction likelihood, with minimal reduction in power output. While PVA models and associated software are often used in conservation research, their use for environmental impact assessment represents a novel and generalizable framework with significant practical value.
BREAKOUT SESSION II, 8:55-9:15 a.m.
A Novel Method For Extracting Toxic Chemotherapeutics from the Human Body
Ashi Gautam, Class of 2016
Research Conducted At: University of California, San Francisco, Department of Radiology and Imaging
Research Mentors: Dr. Daniel Cooke and Dr. Steven Hetts
Faculty Research Mentor: Mr. Christopher Spenner
When chemotherapeutic drugs are injected into a person's body, they spread across the body and have harmful effects due to this spread. The Chemofilter is a filter that contains a resin and is placed in the blood vessel where the chemotherapeutic drug is being injected into the human body. The resin in this filter works to bind with the excess chemotherapeutic, and prevent that excess chemotherapeutic from spreading outside of the targeted region. In this experiment, we determined how to adapt this technology to the cisplatin chemotherapeutic, which is used for head and neck cancer. We tested several resins by performing an absorption assay on these resins and using spectroscopy. After performing these assays on several resins, we conducted a flow model on the Calgon TOG resin, which had the best results in the assays. In this flow model, we used the Masterflex Flow Machine, which is a machine that models blood flow with a pump-tubing structure, to determine how the resin would work with the chemofilter in this setting. Ultimately, the Calgon TOG resin, made of activated carbon, was the resin that binded to the most cisplatin, and thus is the most suitable resin for this experiment.
Force Responsive Reconstruction: Characterizing the Morphogenesis of the Periodontal Ligament through Complementary Biomechanical and Histological Study
Neil Movva, Class of 2015
Research Conducted At: Stanford University
Research Mentor: Dr. Jill A. Helms
Faculty Research Mentor: Ms. Anita Chetty
Siemens Regional Semifinalist; Intel STS Semifinalist; published in Journal of Bone and Mineral Research
Ligament wear injuries impact a large portion of the global human population, ranging from athletes to the elderly. The periodontal ligament (PDL) is a specialized connective tissue, supporting the hardest mineral structure in the human body (enamel/dentin). It has long been of interest for its remarkable self-reconstruction in response to mechanical forces; adapting to maintain a homeostatic morphology. We utilize a novel biomechanical (finite element analysis) and histological approach to predict morphological changes during reconstruction. We demonstrate (unanimously, in five mice per time point) that the PDL responds directly to sites of localized strain, by increasing its own tissue density and collagen fiber attachment patterns while significantly remodeling surrounding bone. Furthermore, we observe that reconstruction is directed in such a way as to restore original force conditions. With our findings, we present a novel hypothesis on the mechanics of ligament reconstruction, holding potentially major implications for clinical reconstructive surgery and future biophysics research.
Coronary Artery Calcification and Cardiovascular Risk Factors in South Asians
Serena Wang, Class of 2015
Research Conducted At: South Asian Heart Center, El Camino Hospital
Research Mentors: Mr. Ashish Mathur, Dr. Cesar Molina
Compared to other ethnic groups, South Asians are at higher risk for cardiovascular disease and diabetes mellitus. Traditional risk factor assessment, developed mainly in a white European-descent populations, may underestimate the incidence of cardiovascular disease in South Asians. Our study examined the relationship between coronary calcification, a strong predictor of cardiovascular events, and other traditional cardiovascular risk factors in South Asians. We analyzed the association of coronary calcification with both traditional and emerging factors commonly used to predict cardiovascular risk. These factors include the lipid panel (total cholesterol, LDL-C, HDL-C, and triglycerides), fasting blood glucose, high-sensitivity CRP, family history of coronary artery disease and diabetes, and the ACC/AHA-recommended atherosclerotic cardiovascular disease (ASCVD) risk score. We found that fasting blood glucose, glycated hemoglobin, insulin, BMI, and personal history of hypertension, hypercholesterolemia and diabetes mellitus were all significantly associated with a non-zero calcium score. High-sensitivity CRP, the conventional lipid panel, the ASCVD risk score, and family history of coronary artery disease and diabetes were not. Because the lipid panel was not predictive while the glucose dysmetabolism risk factors were, the evaluation of South Asians should include and focus on pre-diabetic risk factors, such as fasting blood glucose, insulin, and hemoglobin A1c, and measurements of obesity.
Determining the Ages and Metallicities of Dwarf Elliptical Nuclei and Globular Clusters
Jason Chu, Class of 2015
Research Conducted At: University of California, Santa Cruz, Department of Astronomy and Astrophysics
Research Mentors: Dr. Raja Guhathakurta, Dr. Elisa Toloba
Faculty Research Mentor: Mr. Christopher Spenner
Siemens Regional Finalist; American Astronomical Society Poster Talk; Work included in paper published in Astrophysical Journal
Dwarf elliptical (dE) galaxies are the most common type of galaxy in the universe and function as the building blocks for massive galaxies. Thus, a deeper investigation of dE formation is needed to better understand the origin of large galaxies such as our own Milky Way. One theory of dE formation called dynamical friction proposes that the bright nucleus of many dEs is composed of globular clusters (GCs), or dense families of old stars, that have spiraled into the center of the galaxy. To find evidence of dynamical friction, we used the largest spectroscopic sample of dE nuclei and GCs and employed an algorithm called spectral co-addition to produce a single, composite spectrum with high signal-to-noise ratio. These co-added spectra of dE nuclei and GCs enhanced the clarity of spectral features such as absorption lines and revealed relationships between the objects' age and metallicities. After comparing our co-added spectra with stellar population synthesis model spectra, we determined that the dE nuclei had a metallicity of −0.71 [Fe/H] and an age of 1 Gyr, while the satellite GCs had a metallicity of −1.71 [Fe/H] and an age of 12.59 Gyr. Our findings that dE nuclei are, on average, younger and more metal rich than their satellite GCs establish important benchmarks that dynamical friction and other dE formation theories have to consider.
BREAKOUT SESSION III, 9:20-9:40 a.m.
Characterizing the n-Division Points of Genus-0 Curves Through Straightedge and Compass Constructions
Nitya Mani, Class of 2015
Research Affiliatied With: Stanford University
Research Mentor: Dr. Simon Rubinstein-Salzedo
Siemens Regional Semifinalist
This research project examined three major problems in the field of $n$-division point constructions: determining a bidirectional closed form solution for the regular polygons that can be constructed with a straightedge, compass and trisector; finding the values of $n$ such that the n-division points of the rational $a/b$-hypocycloids can be constructed with compass and straightedge; and seeking a generalization of Abel's Theorem on the Lemniscate to the entire family of Serret curves. We found a closed-form solution for the values of $n$ for which the $n$-division points of a circle can be constructed with a compass, straightedge and trisector, proved a theorem that for all integer $n$, all integer n-division points of any rational $a/b$-hypocycloid are constructible, and determined that with a compass and a straightedge, arbitrary arc lengths on any Serret Curve can be added, subtracted and multiplied.
Towards Rational RNA Therapeutics: 3-D RNA Engineering in a Massive Open Laboratory
Vineet Kosaraju , Class of 2016
Research Conducted At: Stanford University, School of Medicine
Research Mentor: Dr. Rhiju Das
Faculty Research Mentor: Dr. Smriti Koodanjeri
Siemens Regional Finalist
Rationally designed therapeutics based on ribonucleic acid (RNA) molecules have begun to emerge after years of research, with applications to cancer and viral infection. Unfortunately, each drug has taken years to develop due to poor understanding of how RNAs fold to specific secondary structures and then to 3-D structures required for their function. This study aims to uncover the missing design rules using a game-based crowdsourcing approach, the “massive open laboratory." EteRNA has succeeded in recruiting 100,000 players to generate sequences that match a given RNA secondary structure and, through actual experimental feedback, discovering empirically validated design rules. Here, we present a new game EteRNA3D that expands EteRNA from the secondary structure level to design of atomically precise 3-D folds. A pilot study tackles two foundational 3-D problems: stabilizing binding pockets for small molecules to enable external control of RNA therapeutics, and building an RNA “arm" to lock a desired conformation into place. To evaluate success, 12 player and three computer designs have been experimentally synthesized; all were successful in achieving their design targets. Previously unknown rules for 3-D design have already emerged, suggesting that a massive open laboratory will be a powerful paradigm for engineering and testing RNA therapeutics.
A Potential Therapy for Alzheimer's Disease: Encapsulation of Curcumin within Polymeric PLGA-PEG Nanoparticles Protects Neuro2A cells from Beta-Amyloid Induced Cytotoxicity and Improves Bioavailability
Nikash Shankar, Class of 2015
Research Conducted At: University of California, San Francisco, Gladstone Institute of Neurological Disease and Schmahl Science Workshop
Research Mentors: Dr. Keith Vossel, Ms. Ibtisam Khalaf
Faculty Research Mentor: Mr. Christopher Spenner
Siemens Regional Semifinalist; Intel STS Semifinalist
There is an increased need for a versatile drug with multifunctional properties to treat Alzheimer's disease (AD). Curcumin, a principal curcuminoid of turmeric, has anti-amyloid, anti-apoptotic, and antioxidant activities; however, its water insolubility and poor bioavailability limit its efficacy in AD. Nanoparticle-based drug delivery circumvents the pitfalls of curcumin's poor solubility. The goal of this project is to use a polymeric nanoparticle (PEG-PLGA) encapsulated curcumin (nanocurcumin) as an effective vehicle for curcumin delivery to an in vitro AD neuronal cell model and to study its anti-apoptotic and anti-amyloid effects. The results from this study demonstrated that nanocurcumin reduced the βA levels by around 30 percent, exhibited an antioxidant activity of 90 percent, and increased cell viability by over 20 percent when compared to curcumin. Moreover, the 24-hour in vitro release kinetics of nanocurcumin demonstrated a 70 percent curcumin release from nanoparticles, and the fluorescence microscopy images showed that nanocurcumin had a higher cellular uptake than did curcumin. This study is the first to use a polymeric nanoparticle vehicle for curcumin delivery in a cell model of AD. By effectively encapsulating curcumin in PEG-PLGA nanoparticles without destroying curcumin's inherent properties, nanocurcumin with its multiple functions, may thus be a viable potential treatment for AD.
Analyzing First-Trimester MicroRNA as a Marker for Assessing Adverse Pregnancy Risk
Roshni Pankhaniya, Class of 2015
Research Conducted At: Stanford University, Human Immune Monitoring Center Department of Microbiology and Immunology
Research mentors: Dr. Holden Maecker, Dr. Xuhuai Ji
Siemens Regional Semifinalist; Co-authored on paper being submitted to Journal of Reproductive Immunology
The recent discovery of microRNA and the role it plays in regulation of cell pathways, gene expression, and development suggests its potential as a marker for preeclampsia. The latest time for preventative treatment for preeclampsia is prior to 10 weeks gestation; however existing biomarkers are only known to diagnose preeclampsia after 13 weeks gestation. In this research, we validate a method that uses first-trimester maternal cell microRNA to predict adverse pregnancy risk. Our experimental design focused on isolating total RNA and quantifying the activity levels of microRNA through a biomark analysis to gauge for differences in expression in healthy vs. high-risk patients. We made our predictions of adverse pregnancy risk based on the relative strength of expression and compared our predictions to the clinical diagnosis of 46 patients. Our results showed that quantification of miRNA expression in maternal cells could be used to accurately predict adverse pregnancy risk and validated in further detail previous work done in our lab. Furthermore, we identified five microRNAs whose presence was most significant to prediction of adverse pregnancy risk. Though a previous study in our lab did conclude that miRNA could be helpful in diagnosing adverse pregnancy outcome, to our knowledge our study is the first to specifically identify five miRNAs as the most important in determining preeclampsia risk. With the ability to detect preeclampsia prior to 10-week gestation through the use of five specific microRNAs, researchers can began to identify areas for preventative treatment and target cell pathways regulated by these five miRNAs to correct for the causes of preeclampsia.
BREAKOUT SESSION IV, 10-10:20 a.m.
Computer-Aided Genomic Characterization of Colorectal Cancer Driver Alterations for Oncogenic Transformation of Primary Colon Organoids
Steven Wang, Class of 2015
Research Conducted At: Stanford University, Department of Hematology
Research Mentors: Dr. Michael Cantrell, Dr. Calvin Kuo
Intel Science Talent Search Finalist
Understanding the genomic basis of colorectal tumorigenesis is crucial for future diagnostic, prognostic and drug development efforts. Two challenges limit our ability to characterize the cancer genome: unilateral computational methods lack the ability to robustly distinguish “driver" alterations from random events, and current in vitro validation models do not accurately recapitulate the microenvironment of a developing tumor. In this study, we developed a multidimensional framework to identify driver aberrations and validate their tumorigenic ability in colon organoids. By integrating copy number data, pathway analysis, gene expression and survival correlation, we confirmed well-known alterations (KRAS, PIK3CA, APC, RB1 and P53) and discovered novel putative oncogenes (MUC20, RAF1, IKBKB) and tumor suppressor genes (CDC73, PTEN, WWOX). Two well-known aberrations from our analysis, Kras G12D mutation and APC knockout, and a two-hit model of Kras G12D mutation and p53 knockout were validated through Ad-Cre mediated adenoviral infection of colon organoids in an air-liquid interface environment. Significant expansion of infected organoids confirmed tumorigenic properties, implicating the organoid model as a highly amenable system for oncogene validation and tumor engineering. Our innovative computational framework with successful validation in technically advanced colon organoids provides keen insight into colorectal tumorigenesis and holds great promise for applications in oncogene discovery, prognostics and targeted therapy.
BREAKOUT SESSION V, 11:40 a.m.-12 p.m.
A Novel Design and Evaluation of Chitosan Nanoparticle Ocular Drug Delivery System
Sriram Somasundaram, Class of 2015
Research Conducted At: Stanford Immunology and Neuroscience/Advanced Drug Delivery
Research Mentors: Dr. Mohammed Inayathullah, Dr. Wei Jiang
Faculty Research Mentors: Dr. Mala Raghavan, Dr. Gary Blickenstaff
Intel Science Talent Search semifinalist
There is a tremendous lack of ocular drug delivery systems currently, when ocular diseases cause preventable blindness in large developing populations. An eyedrop enhanced with a sufficient drug delivery system could be applied instead of surgery and bedstay. This study takes an interdisciplinary approach combining protein docking models, biophysics test, in vitro system characterization, and pH-dependent permeation to design a novel delivery system. Chitosan fluorescein nanoparticles (C-F-NP) were created and compatible with drugs for the major ocular diseases: cataracts, glaucoma and conjunctivitis. C-F-NP loads drugs on its amino groups and has a pH-dependent release. Additionally, C-F-NP binds best to the cornea at low pH and releases more drug at higher pH's. A pH switch was conducted where the C-F-NP was applied to an in vitro simulated cornea at low pH followed by an increase to high pH. The pH switch resulted in high bioavailability (93.0%) that had the combined effect of both pH's and fivefold increase over the drug without C-F-NP. The novel approach of changing the pH upon administration can be used in other delivery systems. Since the C-F-NP system is inexpensive, biodegradable, and no longer has a corneal permeation drawback, it can combat preventable blindness in the world.
Using Text Mining to Increase the Effectiveness of Online Cancer Support Groups
Ankita Pannu, Class of 2015
Research Conducted At: IBM Almaden Research Center
Research Mentor: Mr. Shankar Venkataraman
Siemens Regional Semifinalist; ACM Bay Area - Top 5 Math/CS Research Project; Presenter at Grace Hopper Conference
Online cancer groups are invaluable platforms for patients to seek medical information as well as emotional support. Patients use these forums to share details about their conditions, supplementing the advice they receive from their doctors. Given that some cancers are on the rise, and novel therapies are being introduced all the time, such groups have become warehouses of real-world patient experiences. Mining these groups presents unique opportunities for patients and healthcare professionals alike. In this project, we exploit natural language processing and text mining techniques to dig into free-form discussion threads that are infused with medical jargon. We start with topic modeling to understand the overall structure of the discussions. Using correlations and association rule mining, we discover useful relationships pertaining to the cancer experience, especially around symptoms and treatments. Lastly, we develop techniques to extract commonly voiced opinions from long discussion threads in order to quickly discover key information and ascertain thread relevancy. We also propose ways of augmenting support group user-interfaces to enhance discoverability of information. We demonstrate these techniques on a real-world thyroid cancer support group.
Physical Properties and Evolution of Gravitationally Bound Halo Structures in Cosmological Dark Matter Simulations
David Lin, Class of 2015
Research Conducted At: University of California, Santa Cruz, Department of Physics and Astronomy
Research Mentors: Dr. Miguel Rocha, Professor Joel R. Primack
Faculty Research Mentor: Mr. Christopher Spenner
Intel STS Semifinalist; AAS Conference Poster Presenter
Dark matter halos existing around visible galaxies are essential to studies of galaxy formation and evolution. Since dark matter cannot be observed directly, studies of dark matter halos are advanced by computer simulations. Normally, halos are defined by their virialized regions; however, regions that are non-virialized can still be gravitationally bound, like the collision-bound Milky Way and Andromeda galaxies. Our project is the first characterization of gravitationally bound halo structures, their properties and their evolution. This study found the larger bound regions surrounding every dark matter halo from a 100 Mpc cube of the Bolshoi Simulation at redshifts 0, 1 and 2. We optimized computation by removing subhalos, implementing a search radius, and parallelizing code across 160 supercomputer cores. Then, we created a mass function, circular velocity function, and correlation function to describe the regions. The evolution of the regions' physical properties was consistent with predictions from a ΛCDM universe model. We characterized the sizes and shapes of these bound regions across different mass intervals and redshifts. Most bound regions are elongated, although they become more spheroidal with time. The results enable astronomers to predict how dark matter halos behave in non-virialized regions of space and deepen our understanding of galaxy formation.
Shedding Light on Human Evolution: Machine Learning Algorithms for Systematic Genome-Wide Discovery and Characterization of Adaptive Mutations
Andrew Jin, Class of 2015
Research Conducted At: Harvard University and the Broad Institute
Research Mentors: Mr. Joseph Vitti, Dr. Daniel Park, Professor Pardis Sabeti
Faculty Research Mentors: Mr. Chris Spenner and Mr. Mike Pistacchi
Intel Science Talent Search National First-Place Winner; Siemens Regional Semifinalist
Positive natural selection has played a central role in the genesis and progress of our species, yet few adaptive mutations and traits are known. In this study, we developed a powerful ensemble machine learning approach – integrating support vector machines, random forests, neural networks and LASSO regression – to systematically probe the entire human genome and pinpoint exact single nucleotide polymorphisms (SNPs) under selection. After training models on millions of simulated SNPs and assessing their prediction accuracy on both simulations and real-life data, we observed that they significantly outperform the current state-of-the-art algorithm: Grossman et al.'s composite of multiple signals (CMS) test. Our machine learning models had nearly perfect AUROC values (>0.99), correctly identified 95-97 percent of selected SNPs, and returned eight times fewer false positives than CMS (which achieved <90 percent sensitivity). The models were then applied to 179 empirical full genome sequences from the 1000G Project to discover new adaptive SNPs, and extensive functional annotation of the results with large-scale genomics datasets revealed 130 exciting candidates linked to potentially beneficial phenotypes. For example, this study was the first to identify putative selected SNPs that decrease schizophrenia and diabetes risk; a nonsynonymous substitution in MOGS possibly confers resistance to viruses such as influenza or hepatitis; and molecular docking was performed to demonstrate that two mutations in the GPATCH1 receptor likely reduce bacterial meningitis susceptibility by obstructing binding with E. coli's outer membrane protein A (OmpA).
Immunomodulation by Human Retinal Pigment Epithelial Cell Line ARPE-19
Shikhar Dixit 2015
Research Conducted At: Boston University
Research Mentor: Dr. Andrew W. Taylor
Faculty Research Mentor: Mr. Jeff Sutton
Intel Science Talent Search Semifinalist; Research Published in Opthamology Journal
The majority of current immunobiological experiments use primary retinal pigment epithelial (RPE) cells to study ocular immunology. The purpose of these experiments is to analyze the effects of the human RPE cell line ARPE-19 (ATCC® CRL¬2302TM) on the activation of phagolysosomes in macrophages that phagocytize bioparticles. This will allow us to know if it is possible to use the more readily available ARPE-19 cells to study ocular immunobiology. This was done by treating macrophages with conditioned media from confluent, sub-confluent, or post-confluent ARPE-19 monolayers, and then examining the cultures under fluorescent microscopy to analyze the suppression of activation of phagolysosome. Results show that macrophages treated with conditioned media from confluent or sub-confluent ARPE-19 cells had no statistical difference in their relative fluorescence intensity from untreated macrophages. In contrast, macrophages treated with conditioned media from three day confluent monolayer ARPE-19 cell cultures were significantly suppressed in their relative fluorescence intensity. The results demonstrated that ARPE-19 cells produced soluble factors that suppressed the activation of phagolysosomes in macrophages, but only when the ARPE-19 cells were an established confluent monolayer. This means that ARPE-19 cells could regulate immunity like primary RPE, and be used in testing new drugs for ocular inflammatory and autoimmune diseases.
BREAKOUT SESSION VI, 3:15-3:35 p.m.
Network Based Integration of High-Throughput Gene Expression and Methylation Data Reveals New Insights into NAFLD Progression
Rohith Kuditipudi, Class of 2015
Research Conducted At: University of California, San Diego, Department of Bioengineering/Systems Biology
Research mentors: Dr. Shankar Subramaniam, Dr. Merril Gersten
Intel Science Talent Search finalist
The increasingly widespread availability of high-throughput genomic data over the course of the past decade has afforded new opportunities for integrating multiple sources of genomic data to reveal new insights into widespread diseases. In particular, non-alcoholic fatty liver disease (NAFLD) is currently estimated to affect as much as 30 percent of the American population. Although many with NAFLD experience minimal discomfort and can typically continue to lead normal lives, some individuals soon progress to non-alcoholic steatohepatitis (NASH), resulting in severe hepatic inflammation and the potential onset of liver cirrhosis and even hepatocellular carcinoma (HCC). Currently, relatively little is known regarding the underlying mechanisms driving this progression, and it is nearly impossible to predict an individual's susceptibility to NASH. However, in the past few years, an increasing number of studies have by association generally implicated DNA methylation as a potential driver of progression. The present study employs a data-driven network based approach to identify pathways linked to disease progression and likely affected by methylomic trends. Weighted gene co-expression network analysis is implemented to identify modules of highly connected genes, all either differentially expressed or differentially methylated in NASH, which remain preserved in both the expression and methylation data. Canonical correlation analysis is then leveraged to quantify trends among the DNA methylation profiles of genes within sub-networks of interest. This approach yielded meaningful, novel insights regarding pan-omic interactions that may drive the progression from NAFLD to NASH, including in several previously overlooked pathways with direct functional relation to NASH.
Characterizing Novel Binders as Tools for Understanding Chloride Transport Mechanisms
Cindy Liu, Class of 2015
Research Conducted At: Stanford University, Department of Molecular and Cellular Physiology
Research Mentors: Christian Evans, Dr. Merritt Maduke
Faculty Research Mentor: Mr. Christopher Spenner
Intel STS Semifinalist
CLCs are chloride/proton antiporters that transport chloride ions and protons across membranes. Dysfunctional chloride channels have been linked to diseases such as epilepsy and myotonia. High-resolution crystal structures of CLC proteins have provided understanding of their mechanisms in health and disease. However, structures are available for only one CLC conformational state. We seek to develop "molecular chaperones" that can facilitate crystallization and structure-determination of CLC. To this end, a library of synthetic antigen binders ("monobodies") was screened for binders specific to CLC-ec1, a prokaryotic CLC homolog previously crystallized in one conformational state. Six high-affinity binders were identified. To identify potential "conformation specific" binders, we developed an assay to determine the effect of the monobodies on CLC function, reasoning that binding to a specific conformation will "trap" that conformation and inhibit activity. In this project, two monobodies, 3L6 and 4L7, were overexpressed, purified, and tested. Both inhibit CLC-ec1, with IC50 values of 1.1 µM ± 0.02 µM and 7.7 µM ± 3.8µM respectively. Ongoing studies involve experiments to determine whether 3L6 and 4L7 bind specifically to certain conformations, with follow up studies planned to crystallize 3L6 and/or 4L7 bound to CLC-ec1. The novel binders discovered may be useful as lead compounds in therapies to treat CLC-related diseases.
Tracking Parallel Mutation Trajectories Conferring Increased Resistance to HIV-RT Inhibitors
Sahana Rangarajan, 2015
Research Conducted At: University of California, Santa Cruz, Microbiology and Environmental Toxicology Department
Research mentors: Mr. Jay Kim, Dr. Manel Camps
Many HIV drugs target reverse transcriptase (RT), an enzyme essential to the virus' genome replication and incorporation into host cells. However, due to HIV's high rate of mutation, strains develop and accumulate mutations that confer drug resistance. In our study, we used longitudinal patient data to examine the relationships between known drug resistance mutations. We first used the Jaccard coefficient and permutation test for significance in order to determine the strength of dependencies between significant pairs of mutations. Through an undirected co-occurrence network, we found that HIV-RT resistance mutations split into two main subcategories based on whether they confer drug resistance or increase fitness in some other way. Within the drug resistance subcommunity, we found that they cluster additionally based on category of drug resistance. In our directed Bayesian network, we found that known mutation trajectories are represented, increasing the viability of longitudinal data as a way to predict drug resistance mutations and mutation trajectories that arise in patients after drugs are administered. Our future direction involves mainly improving our ability to predict HIVRT drug resistance using multidimensional Bayesian networks, the use of phenotypic data, analysis of patient-specific subnetworks, and factoring amino acid changes into studying the mutations.
Evaluation of Novel Mechanisms of Heart Failure in Children with Congenital Heart Disease Identifies the Potential for Pediatric-Specific Heart Failure Therapies
Rahul Jayaraman, Class of 2015
Research Conducted At: Stanford University, Department of Pediatrics
Research mentors: Dr. Sushma Reddy, Dr. Dong-Qing Hu
The hypertrophied right ventricle (RV) is uniquely at risk for heart failure in children with cardiac birth defects such as tetralogy of Fallot, L-TGA, and hypoplastic left heart syndrome. Differences in pathways between the left and right ventricles regulating oxidant stress, apoptosis, and angiogenesis predispose the RV to a more rapid progression to failure. We hypothesized that the RV angiogenic response is impaired in RV hypertrophy (RVH) and evaluated the mechanisms of this attenuated response, in particular the role of endothelial cell (EC) microRNAs (miRs). We created pulmonary stenosis (PS) to induce RVH in adult male FVB mice. The hearts were harvested after ten days, and RNA and protein isolated from the RV free wall. Angiogenic pathways mediated by Hif-1α, VEGFs, and SIRT1 were assessed by Western blotting. The expression of endothelial miRs 34a and 148a was assessed by qRT-PCR and compared to sham-operated controls, and the capillary-myocyte ratio was assessed by CD31 staining of frozen sections. Mice developed RVH by ten days, at which time the capillary-myocyte ratio was decreased in PS vs. sham-operated controls (1.26±0.02 vs. 1.58±0.05, p<0.05). This was accompanied by increased nuclear Hif-1α expression (1.51±0.38 vs. 0.29±0.07, p<0.05) and decreased VEGF-A (1.68±0.24 vs. 3.11±0.18, p<0.05), VEGF-B (0.84±0.01 vs. 2.63±0.03, p<0.05), and SIRT1 (0.11±0.01 vs. 0.42±0.12, p=0.07) expression. In addition, miRs 34a and 148a were upregulated in RVH by 6.5±2 fold and 1.6±0.3 fold, respectively (p<0.05). In summary, we have identified impaired VEGF signaling despite activation of the upstream master regulator of angiogenesis, Hif-1α, as a potential mechanism for the decreased capillarity seen in RVH/RVF. Interestingly, miRs known to mediate EC proliferation and survival via VEGFs and SIRT1 were upregulated, suggesting an association between these miRs and the attenuated angiogenesis in RVH/RVF. Overexpression of these miRs may predispose the stressed RV to RVF. Inhibition of these miRs may be used as an RV-specific therapeutic strategy to restore angiogenesis and thereby slow the progression to RVF.
Improving Concentrated Solar Power Technologies: Planar Optical Waveguides for Transporting Concentrated Light to Enable Efficient Energy Conversion
Kailas Vodrahalli, Class of 2015
Research Conducted At: University of California, Santa Cruz, NASA Ames
Research mentor: Professor Nobuhiko Kobayashi
Intel Science Talent Search semifinalist
Solar energy is the cleanest, most abundant form of green energy available; however, it is dilute and concentrating solar power is beneficial. New technologies like concentrated solar power systems (CSPs), concentrated photovoltaics (CPV) and daylighting are promising, but improvements are needed. Transporting concentrated solar power through optical fibers eliminates transporting hot fluids in CSPs, and improves efficiencies in CPVs and daylighting applications; however, an optical coupler that can focus, transport and couple multimode solar light into the fibers is required. Through computer modeling, this study evaluates the viability of high refractive index planar waveguides for this optical coupler application. Characterization studies performed identified key design requirements for the waveguide: refractive index ≥ 1.62, extinction coefficient < 1E-7, and taper angle ≤ 6 degrees for > 99 percent power transmission of multimode light. Planar waveguides have been successfully shown to focus and transmit multimode light without requiring any complex grating or graded index structures for the first time. A novel multi-taper structure is proposed and it has a remarkable > 99 percent transmission (about a 43 percent improvement over an alternate fiber-coupler solution). This research represents a significant step forward in the development of optical couplers for improving concentrated solar energy systems to provide clean, abundant and affordable energy.