Roger Lefort's headshot

Roger Lefort, PhD
Data Scientist | Project Manager

ABOUT ME


Summary

Data scientist and Neuroscientist with strong statistics and data analytics background and broad-based experience in analyzing large datasets and using predictive modeling, data processing, and data mining algorithms to come up with data-driven strategies to solve complex problems. Enthusiastic team-player and communicator and founder of several organizations and clubs. Extensive leadership and mentoring experience developed from 10+ years of building and managing cross-functional teams and projects.

Learning is what drives me — I believe that there is always room for improvement, especially with regards to my own abilities, and I constantly seek to sharpen my skills.


TIMELINE

My career path so far... Click on the timeline for more datails.


SKILLS

brainNeuroscience

I have over a decade of experience in CNS disease research, with a specific focus on Alzheimer's disease. I specialize in in vivo, in vitro and ex vivo disease modeling, using a variety of tools, including transgnenic mice and cell-derived models (primary and ESC/iPSC). I am highly skilled in biochemistry and molecular biology techniques, and data analysis and interpretation. I also have deep knowledge of the scientific literature and molecular signaling.

chartData Science
I am passionate about data and story-telling through data, and I don't shy away from messy datasets! I care deeply about things like statistical significance and reproducibility. I am no graphic artist, but I like to draw stuff, make pretty graphs and presentations, and mess with colors and line widths. Data is no good without visualization.
promptProgramming
My language of choice is Python, and its numerous data science libraries (pandas, numpy, scipy, keras, scikit, etc). I can also work with HTML/CSS, Javascript and SQL. My skills are more adapted for rapid prototyping and data analysis than production quality code. By the way, this resume is built using HTML, Bootstrap, CSS and Javascript D3, served through a flask server.

Other skills


Experience

Education

Continuing Education


Select Publications

research

Effects of Eph/ephrin signalling and human Alzheimer's disease-associated EphA1 on Drosophila behaviour and neurophysiology.

Buhl E, Kim YA, Parsons T, Zhu B, Santa-Maria I, Lefort R, Hodge JJL.

Neurobiol Dis. 2022 Aug;170:105752

lzheimer's disease (AD) is the most prevalent neurodegenerative disease placing a great burden on people living with it, carers and society. Yet, the underlying patho-mechanisms remain unknown and treatments limited. To better understand the molecular changes associated with AD, genome-wide association studies (GWAS) have identified hundreds of candidate genes linked to the disease, like the receptor tyrosine kinase EphA1. However, demonstration of whether and how these genes cause pathology is largely lacking. Here, utilising fly genetics, we generated the first Drosophila model of human wild-type and P460L mutant EphA1 and tested the effects of Eph/ephrin signalling on AD-relevant behaviour and neurophysiology. We show that EphA1 mis-expression did not cause neurodegeneration, shorten lifespan or affect memory but flies mis-expressing the wild-type or mutant receptor were hyper-aroused, had reduced sleep, a stronger circadian rhythm and increased clock neuron activity and excitability. Over-expression of endogenous fly Eph and RNAi-mediated knock-down of Eph and its ligand ephrin affected sleep architecture and neurophysiology. Eph over-expression led to stronger circadian morning anticipation while ephrin knock-down impaired memory. A dominant negative form of the GTPase Rho1, a potential intracellular effector of Eph, led to hyper-aroused flies, memory impairment, less anticipatory behaviour and neurophysiological changes. Our results demonstrate a role of Eph/ephrin signalling in a range of behaviours affected in AD. This presents a starting point for studies into the underlying mechanisms of AD including interactions with other AD-associated genes, like Rho1, Ankyrin, Tau and APP with the potential to identify new targets for treatment.

Abstract »

research

Type 1 Interleukin-4 Signaling Obliterates Mouse Astroglia in vivo but Not in vitro.

Mashkaryan V, Siddiqui T, Popova S, Cosacak MI, Bhattarai P, Brandt K, Govindarajan M, Petzold P, Reinhardt S, Dahl A, Lefort R, Kizil C.

Front Cell Dev Biol. 2020 Feb 26;8(114)

Recent findings suggest that reduced neurogenesis could be one of the underlying reasons for the exacerbated neuropathology in humans, thus restoring the neural stem cell proliferation and neurogenesis could help to circumvent some pathological aspects of Alzheimer's disease. We recently identified Interleukin-4/STAT6 signaling as a neuron-glia crosstalk mechanism that enables glial proliferation and neurogenesis in adult zebrafish brain and 3D cultures of human astroglia, which manifest neurogenic properties. In this study, by using single cell sequencing in the APP/PS1dE9 mouse model of AD, we found that IL4 receptor (Il4r) is not expressed in mouse astroglia and IL4 signaling is not active in these cells. We tested whether activating IL4/STAT6 signaling would enhance cell proliferation and neurogenesis in healthy and disease conditions. Lentivirus-mediated expression of IL4R or constitutively active STAT6VT impaired the survival capacity of mouse astroglia in vivo but not in vitro. These results suggest that the adult mouse brain generates a non-permissive environment that dictates a negative effect of IL4 signaling on astroglial survival and neurogenic properties in contrast to zebrafish brains and in vitro mammalian cell cultures. Our findings that IL4R signaling in dentate gyrus (DG) of adult mouse brain impinges on the survival of DG cells implicate an evolutionary mechanism that might underlie the loss of neuroregenerative ability of the brain, which might be utilized for basic and clinical aspects for neurodegenerative diseases.

Abstract »

research

Alzheimer’s disease-associated P460L mutation in ephrin receptor type A1 (EphA1) leads to dysregulated Rho-GTPase signaling.

Kim Y, Lasso G, Patel H, Vardarajan B, Santa-Maria I, Lefort R.

bioRxiv 2021.06.17.448790

Recently, late onset AD (LOAD) genome-wide association studies identified EphA1, a member of receptor tyrosine kinase family (RTK) as a disease associated loci. In the follow-up study where 3 independent LOAD cohorts were performed, a P460L coding mutation in EphA1 loci showed a significant association with LOAD. However, the role of EphA1 and P460L mutant EphA1 in AD is not fully understood. We have characterized this mutation biophysically and biochemically. Our structural in silico model and in vitro biochemical analysis demonstrate that EphA1-P460L mutation makes the receptor constitutively active suggesting a gain-of-toxic function leading to chronic EphA1 signaling in the brain. Moreover, we report that the EphA1 P460L variant triggers Rho-GTPase signaling dysregulation that could potentially contribute to spine morphology abnormalities and synaptic dysfunction observed in AD pathology.

Abstract »

research

Dysfunction of the ubiquitin ligase E3A Ube3A/E6-AP contributes to synaptic pathology in Alzheimer’s disease.

Olabarria M, Pasini S, Corona C, Robador P, Song C, Patel H, Lefort R.

Commun Biol. 2019 Mar 22;2(111)

Synaptic dysfunction and synapse loss are prominent features in Alzheimer's disease. Members of the Rho-family of guanosine triphosphatases, specifically RhoA, and the synaptic protein Arc are implicated in these pathogenic processes. They share a common regulatory molecule, the E3 ligase Ube3A/E6-AP. Here, we show that Ube3A is reduced in an Alzheimer's disease mouse model, Tg2576 mouse, which overexpresses human APP695 carrying the Swedish mutation, and accumulates Aβ in the brain. Depletion of Ube3A precedes the age-dependent behavioral deficits and loss of dendritic spines in these mice, and results from a decrease in solubility following phosphorylation by c-Abl, after Aβ exposure. Loss of Ube3A triggers the accumulation of Arc and Ephexin-5, driving internalization of GluR1, and activation of RhoA, respectively, culminating in pruning of synapses, which is blocked by restoring Ube3A. Taken together, our results place Ube3A as a critical player in Alzheimer's disease pathogenesis, and as a potential therapeutic target.

Abstract »

research

Assembly and interrogation of Alzheimer's disease genetic networks reveal novel regulators of progression.

Aubry S, Shin W, Crary JF, Lefort R, Qureshi YH, Lefebvre C, Califano A, Shelanski ML.

PLoS One. 2015 Mar 17;10(3):e0120352

Alzheimer's disease (AD) is a complex multifactorial disorder with poorly characterized pathogenesis. Our understanding of this disease would thus benefit from an approach that addresses this complexity by elucidating the regulatory networks that are dysregulated in the neural compartment of AD patients, across distinct brain regions. Here, we use a Systems Biology (SB) approach, which has been highly successful in the dissection of cancer related phenotypes, to reverse engineer the transcriptional regulation layer of human neuronal cells and interrogate it to infer candidate Master Regulators (MRs) responsible for disease progression. Analysis of gene expression profiles from laser-captured neurons from AD and controls subjects, using the Algorithm for the Reconstruction of Accurate Cellular Networks (ARACNe), yielded an interactome consisting of 488,353 transcription-factor/target interactions. Interrogation of this interactome, using the Master Regulator INference algorithm (MARINa), identified an unbiased set of candidate MRs causally responsible for regulating the transcriptional signature of AD progression. Experimental assays in autopsy-derived human brain tissue showed that three of the top candidate MRs (YY1, p300 and ZMYM3) are indeed biochemically and histopathologically dysregulated in AD brains compared to controls. Our results additionally implicate p53 and loss of acetylation homeostasis in the neurodegenerative process. This study suggests that an integrative, SB approach can be applied to AD and other neurodegenerative diseases, and provide significant novel insight on the disease progression.

Abstract »

research

The Aβ₁₋₄₂ peptide regulates microtubule stability independently of tau.

Pianu B, Lefort R, Thuiliere L, Tabourier E, Bartolini F.

J Cell Sci. 2014 Mar 1;127(Pt 5):1117-27

Interference with microtubule stability by beta-amyloid peptide (Aβ) has been shown to disrupt dendritic function and axonal trafficking, both early events in Alzheimer's disease. However, it is unclear whether Aβ regulation of microtubule dynamics can occur independently of its action on tau. RhoA has been implicated in neurotoxicity by Aβ but the mechanism by which this activation generates cytoskeletal changes is also unclear. We found that oligomeric Aβ1-42 induced the formation of stable detyrosinated microtubules in NIH3T3 cells and this function resulted from the activation of a RhoA-dependent microtubule stabilization pathway regulated by integrin signaling and the formin mDia1. Induction of microtubule stability by Aβ was also initiated by dimerization of APP and required caspase activity, two previously characterized regulators of neurotoxicity downstream of Aβ. Finally, we found that this function was conserved in primary neurons and abolished by Rho inactivation, reinforcing a link between induction of stable detyrosinated microtubules and neuropathogenesis by Aβ. Our study reveals a novel activity of Aβ on the microtubule cytoskeleton that is independent of tau and associated with pathways linked to microtubule stabilization and Aβ-mediated neurotoxicity.

Abstract »

research

Caspase-2 is required for dendritic spine and behavioural alterations in J20 APP transgenic mice.

Pozueta J, Lefort R, Ribe EM, Troy CM, Arancio O, Shelanski M.

Nat Commun. 2013;4:1939

Caspases have critical roles in Alzheimer's disease pathogenesis. Here we show that caspase-2 is required for the cognitive decline seen in human amyloid precursor protein transgenic mice (J20). The age-related changes in behaviour and dendritic spine density observed in these mice are absent when they lack caspase-2, in spite of similar levels of amyloid beta (Aβ) deposition and inflammation. A similar degree of protection is observed in cultured hippocampal neurons lacking caspase-2, which are immune to the synaptotoxic effects of Aβ. Our studies suggest that caspase-2 is a critical mediator in the activation of the RhoA/ROCK-II signalling pathway, leading to the collapse of dendritic spines. We propose that this is controlled by an inactive caspase-2/RhoA/ROCK-II complex localized in dendrites, which dissociates in the presence of Aβ, allowing for their activation and entry in the spine. These findings directly implicate caspase-2 as key driver of synaptic dysfunction in Alzheimer's disease and offer novel therapeutic targets.

Abstract »

research

Cross-linking of cell surface amyloid precursor protein leads to increased β-amyloid peptide production in hippocampal neurons: implications for Alzheimer's disease.

Lefort R, Pozueta J, Shelanski M.

J Neurosci. 2012 Aug 1;32(31):10674-85

The accumulation of the β-amyloid peptide (Aβ) in Alzheimer's disease (AD) is thought to play a causative role in triggering synaptic dysfunction in neurons, leading to their eventual demise through apoptosis. Aβ is produced and secreted upon sequential cleavage of the amyloid precursor protein (APP) by β-secretases and γ-secretases. However, while Aβ levels have been shown to be increased in the brains of AD patients, little is known about how the cleavage of APP and the subsequent generation of Aβ is influenced, or whether the cleavage process changes over time. It has been proposed that Aβ can bind APP and promote amyloidogenic processing of APP, further enhancing Aβ production. Proof of this idea has remained elusive because a clear mechanism has not been identified, and the promiscuous nature of Aβ binding complicates the task of demonstrating the idea. To work around these problems, we used an antibody-mediated approach to bind and cross-link cell-surface APP in cultured rat primary hippocampal neurons. Here we show that cross-linking of APP is sufficient to raise the levels of Aβ in viable neurons with a concomitant increase in the levels of the β-secretase BACE1. This appears to occur as a result of a sorting defect that stems from the caspase-3-mediated inactivation of a key sorting adaptor protein, namely GGA3, which prevents the lysosomal degradation of BACE1. Together, our data suggest the occurrence of a positive pathogenic feedback loop involving Aβ and APP in affected neurons possibly allowing Aβ to spread to nearby healthy neurons.

Abstract »


PORTFOLIO

Click on images for website

National Parks Logo
Strava Dashboard w/ Ploty and Streamlit

This dashboard accesses data through the Strava API to track my activities and workouts. Its purpose is to create a personal activity heatmap (similar to the paid Strava feature for Summit members), and encompasses a greater variety of data analysis and visualization options. A full list of features is listed below:

- Data Updated in realtime and stored on AWS
- Lifetime stats and yearly progression
- Summary table of all activities
- Ride maps and stats + elevation profile from Google Elevation API
- Goals and progress charts
- Bike-specific metrics
- Historical map of all bike rides

Read More »

Card image cap
Analyzing NYC Rat Population w/ Tableau

Rats in New York City are prevalent, as in many densely populated areas. For a long time, the exact number of rats in New York City was unknown, and a common urban legend was that there were up to four times as many rats as people. In 2014, however, scientists more accurately measured the entire city's rat population to be approximately only 25% of the number of humans; i.e., there were approximately 2 million rats to New York's 8.4 million people at the time of the study.
Data is from 2010-Sept 16th, 2017 and includes date, location (lat/lon), type of structure, borough, and community board.
Questions
Q1 - Which Borough and zip codes have the most rat sightings?
Q2 - Is there a correlation with zip code population?
Q3 - Have rat sightings increased over the years?
Q4 - Is there a seasonal pattern to rat sightings?

Read More »

SOTU
NLP - Analyzing State of the Union Speeches

Summary: Scraping transcripts of State of the Union Addresses from The American Presidency Project, using Selenium and BeautifulSoup.

- 229 transcripts and dates were collected from 43 presidents
- Data was augmented with party information and transcripts were cleaned to generate a corpus and a document-term matrix (DTM) with Sklearn
- Exploratory Data Analysis was done on the DTM to generate word clouds and a bar graph of the unique words used by each president during their speeches.
- Sentiment Analysis was done on the cleaned corpus using Spacy and TextBlob to evaluate how speeches have evolved over time and to determine whether these is a clear difference across party lines.

Read More »

NFL Plays
A Machine Learning Approach to Predicting NFL Plays w/ Scikit-Learn

In the National Football League (NFL), being able to accurately predict which play your oppent will call given a set of circumstances would be an invaluable tool, and teams devote many resources towards gaining insights into an opponent's tendencies on the field. One area of immense interest is assessing a team's propensity to run or pass the ball in a given situation, which for a defense can inform critical decision-making about strategy, personnel, and positioning on the field, decisions that have a profound impact on the outcome of a game.
In this project, I take a data-driven approach to classifying offensive play types and examine whether tools from machine learning can offer value in characterizing offensive tendencies.

Read More »

beating heart
Predicting Cardiovascular Disease Risk w/ Scikit-Learn

People with CVD or those who are at high cardiovascular risk (due to the presence of one or more risk factors such as hypertension, diabetes, hyperlipidaemia or other comorbidities) need early detection and management wherein a machine learning model can be of great help.

Read More »

CompVis
Detecting Pathologies w/ Convolutional Neural Networks

This Streamlit-based web application is designed to assist in the medical analysis by automatically detecting and classifying tumors in ultrasound images as benign or malignant. Utilizing advanced computer vision and machine learning techniques, the app provides healthcare professionals with a rapid, initial assessment tool that aids in the diagnosis and understanding of potential pathologies.
Features include:
- Image Upload: Users can upload ultrasound images of tumors for analysis. - Contour Analysis: The app processes uploaded images to detect and highlight contours of tumors, facilitating visual assessment of size and shape. - Sphericity Measurement: By evaluating the roundness of detected contours, the app assists in identifying characteristics typically associated with benign or malignant tumors. - Threshold Adjustment: Users can adjust the threshold settings for contour detection and sphericity to fine-tune the analysis based on specific case needs. - Machine Learning Prediction: The core of the app lies in its ability to predict whether a tumor is likely benign or malignant, using a trained machine learning model. The prediction is accompanied by a probability score, providing insight into the model's confidence. Interactive Visuals: The application includes interactive heatmaps and contour overlays, allowing for an in-depth examination of the tumor's properties.

Read More »


ONLINE PRESENCE

github linkedin twitter ORCID Strava Medium