\( \def\RR{\bf R} \def\real{\mathbb{R}} \def\bold#1{\bf #1} \def\d{\mbox{Cord}} \def\hd{\widehat \mbox{Cord}} \DeclareMathOperator{\cov}{cov} \DeclareMathOperator{\var}{var} \DeclareMathOperator{\cor}{cor} \newcommand{\ac}[1]{\left\{#1\right\}} \DeclareMathOperator{\Ex}{\mathbb{E}} \DeclareMathOperator{\diag}{diag} \)

Complex Modeling of Brain Dynamics


Xi (Rossi) LUO


Brown University
Department of Biostatistics
Center for Statistical Sciences
Computation in Brain and Mind
Brown Institute for Brain Science
Brown Data Science Initiative
The ABCD Research Group


RI NIH IDeA Symposium, Providence, RI
June 2, 2017


Funding: NSF/DMS (BD2K) 1557467; NIH R01EB022911, P20GM103645, P01AA019072, P30AI042853

COBRE CCNSF

  • Core leaders: Jerome Sanes, Sheila Blumstein, John Davenport, Zhijin (Jean) Wu
  • Many project leaders working on diverse problems
    • From flies, to monkeys, and to humans
  • Heterogeneous and complex data
      Genetics, spike trains, optogenetics, electroencephalogram/EEG, functional MRI

Collaborative Team

  • BIBS: Jeff Moher, Dan McCarthy, Maro Machizawa, Joo-Hyun Song
  • Biostat: Yi Zhao, Brendan Le
  • CCV: Peisi Yan
  • UCSF: Steve Gee, Vikaas Sohal
  • UPenn: Dylan Small
  • SHJTU: Weidong Liu
  • Many scientists who made their data publicly available!

$10^{11}$ neurons
Ex: optogenetics modeling Luo et al, Stat Med, 16


$10^4$ genes, $10^6$ SNPs
Ex: Gene networks Liu & Luo, JMVA, 15


$10^6$ functional MRI voxels
Ex: Brain networks Yi & Luo, 17+

  • Overview: machine learning, Bayesian, probability/matrix theory, optimization, large-scale computing
  • Idea: integrate techniques from CS, Biology, Math, and Stat to uncover brain dynamics from data

Ex 1: Genes and HIV Brains

  • Gene expression arrays on post-mortem brain tissues Borjabad et al, 2011
    • Tens of thousands genes
    • Outcome: controls, treated HAND, untreated HAND
  • Data: publicly available from NIH GEO GSE28160
    • R package also publicly available from CRAN
    • Follow NIH's call for reproducible research
  • Question: are there relationships between expressions and HIV treatment?

Data

Value distribution provided by NCBI Portal https://www.ncbi.nlm.nih.gov/geo/geo2r/

Method Comparision

  • Massive testing
    • T tests for each gene vs HAND treatment (Yes/No)
    • Drawbacks: far from biology, no validation, sample size, multiple testing
  • Network modeling
    • Goal: how all genes together as networks related to treatment
    • Need computational and mathematical tools to find gene networks
  • Black-box machine learning and artificial intelligence
    • Feed genes into predictive models
    • Examples: random forests, support vector machine
    • Test prediction performance on an independent sample

Results


Estimated Gene Networks
Our (red line) prediction vs other AIs (higher accuracy better)

Other Applications: fMRI, EEG, ...

  • Input matrix (csv): observations (row) $\times$ variables (col)
  • Output: network, prediction (under construction)


DEMO

Ex 2: Optogenetics

  • Technology to stimulate neurons
  • Nature Method of the Year 2010

Data ( light stimuli, neural spikes)

Model

Prediction Comparision

Our PRO/PROs models have higher accuracy

Ex 3: EEG and Behavior

  • Change of mind (CoM) by hand movement trajectory
  • Can we predict CoM before the movement?

Our model predicts CoM using EEG before movement starts (definitely before CoM happens)

Ex 4: Beyond "Simple" Networks

Question: quantify red, blue, and other pathways
from stimulus to orange outcome circle/region Heim et al, 09

Pathway=Activation+Connectivity

  • Activation: stimulus $\rightarrow$ brain region activity
  • Connectivity: one brain region $\rightarrow$ another region
    • Whether not two or more brain regions "correlate"
  • Pathway: stimulus $\rightarrow$ brain region A $\rightarrow$ region B
  • Strong path: strong activation and strong conn
  • Zero path: zero activation or zero conn, including
    • Zero activation + strong conn = zero
    • Strong activation + zero conn = zero

Stim-M25-R and Stim-M65-R significant shown largest weight areas

  • M65 responsible for language processing, larger flow under story
  • M25 responsible for uncertainty, larger flow under math

Thank you!


Comments? Questions?


Slides at: BrainDataScience.com

or BigComplexData.com