OPT-PDE and FRET Imaging of Molecular Transport in Live Cells

Name of the University

NAME

NO:

Faculty

Degree: Master Thesis

Referee:

Supervisor:

NAME

NO:

January, 2017

1
TABLE OF CONTENTS

List of Tables ii

List of Figures iii

1 Introduction 1

1.1 Background to the Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1.1 The Control-State-Operator . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.1.2 Existence and Uniqueness of Optimal Control Solutions . . . . . . . . . . 3

1.1.3 Subdifferential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.1.4 Necessity Condition and Optimality System . . . . . . . . . . . . . . . . . 4

1.1.5 Construction of optimal control solutions . . . . . . . . . . . . . . . . . . 5

1.1.6 Convergence of the solutions . . . . . . . . . . . . . . . . . . . . . . . . . 5

1.2 Preliminary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.3 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.4 Research Question . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2 LITERATURE REVIEW 8

2.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

3 Methodology 9

4 Results and Discussion 11

4.1 Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

REFERENCES 13

i
List of Tables

ii
List of Figures

1 Sub-differential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

iii
Declaration of Authorship

I, , declare that this thesis titled, OPT-PDE and FRET Imaging of Molecular Transport

in Live Cells and the work presented in it are my own. I confirm that all information in

this document has been obtained and presented in accordance with academic rules. I have

acknowledged all main sources of help.

Signed:

Date:

iv
Abstract

The paper entails development of the forward simulation system in live-cell image analysis soft-

ware package Fluocell. The paper also focuses on building an interface between Fluocell and Dr.

Banks OPTPDE solver software package PLTMG besides developing and characterize OPT-PDE

solvers simulations generated in Fluocell. The final task will be to pre-process live-cell images in

Fluocell and use PLTMG to reconstruct diffusion maps for the pictures.

v
Chapter 1

Introduction

1.1 Background to the Study

Molecular diffusion and transport are fundamental physical processes in biology and life. These

processes are essential for the biological functions in cell and tissue, manifested by molecular

transport and interaction in live cells, the propagation of cardiac action potential in the heart,

and the regulation of morphogen gradient during development. The diffusion or transport of a

given molecule can be mainly related to its activity or function. However, it remains an enigma

regarding whether or how the molecular activity is causally and quantitatively coupled to the

transport process at subcellular regions in live cells.

Posing and answering of these fundamental questions remains a tremendous challenge due to a

lack of quantitative methods to reconstruct the subcellular diffusion map based on imaging data,

and visualize subcellular molecular activities and evaluate the correlation between molecular

activity and transport. While the cross-correlation-based FCS methods are easy to implement

for analyzing the diffusion process in live cells, only images of low molecular density can be

compatible to result in accurate outcomes. Furthermore, the spatial resolution for FCS remains

relatively low [1]. In contrast, the model-based methods such as finite difference or finite ele-

ment methods have the advantage of being theoretically well-defined, with high spatio-temporal

resolution and less limitation on molecular density [5]. In particular, optimization models with

1
partial differential equation constraints (OPT-PDEs) are incredibly versatile and ideally suited

for these applications. The OPT-PDE analysis method for constructing diffusion tensor map

in live cells is a significant extension from current models and applications. This method can

be utilized to construct diffusion tensor maps for any molecule of interest from cellular images,

which so far has only been achieved for some small molecules with raster FCS methods [3, 8].

Consequently, the project will concentrate on development of an optimization and finite-element

based diffusion analysis method aims at re-constructing spatio-temporal diffusion maps from

live-cell images. The diffusion analysis method will be characterized and tested with computer-

generated simulations with known diffusion maps. In this project, we will formulate and char-

acterize an optimization model with PDE constraints (OP T P DE) for computing the spatio-

temporal diffusion tensor maps of any given intracellular molecule. The corresponding large

scale linear system and solver algorithms will be developed and implemented with regards to

large scale linear system and solver algorithm, and investigate the convergence property and

accuracy of the solver.

In an attempt to attain the objective set, the paper will entails integration of leading-edge mathe-

matical science with quantitative live cell imaging, to provide a general and systematic approach

to reconstruct diffusion maps as well as maps of constitutive relations between physical and chem-

ical properties of functional molecules with high spatio-temporal resolution. The paper motivates

development of advanced mathematical theory and computational algorithms, whose application

fields are anticipated in computational biology, cell biology, biophysics and biochemistry.

The following subsection outline the major definition of the basic theoretical aspect that must be

understood before understanding how the choice of regularizing functions affect the optimality

of spatio-temporal solutions for the diffusion equation. The subsection also show the outline of

the thesis.

1.1.1 The Control-State-Operator

If (x , t ) are the solution of the Navier-Stokes systems as portrayed by [2] with w bing on the

right hand side, and v > M(t ), then the control-to-state operator G : L2 () U is singled-

2
valued near w and Frechet differentiable at w . Furthermore, the derivative of the systems at

w , in the direction v, shall be given by the unique solution t of the system:

vt + (x.)x + (x .)x + t = v in

div x = 0 in (1)

x| = 0 on .

1.1.2 Existence and Uniqueness of Optimal Control Solutions

If

x = Ax + Bu, (2)

is a system with general linear time-invariant dynamics, with x n and u U m . If the

control objective is set from the initial state given by x(t0 ) = x0 to a given final state x1 in

minimal time and assuming that x1 Rt (x0 ) for some t t0 , then there exist a time optimal

control solution of equation (2) [4].

If equation (2) is c.c. and the pair (A, Q) is c.o., then the control which minimizes

Z
(xT Qx + uT T u)dt, (3)
0

is given by

u (t) = R1 B T P x(t), (4)

where P is unique positive definite symmetric matrix which satisfies,

P BR1 B T P AT P P A Q = 0, (5)

3
whose positive definite solution yields a unique solution of equation (2) [6].

1.1.3 Subdifferential

Consider the figure below:

Figure 1: Sub-differential

Then, n is said to be a sub-differential of v at x if

v(y) (x) + h, y xi o(|y x|). (6)

Therefore, the set of a sub-differential of v is denoted by D v(x) [4].

1.1.4 Necessity Condition and Optimality System

Consider the following problem:

Z ZZ
min J(y, v, u) := (x, y(x, T )) dx + (x, t, y(x, t), v(x, t)) dx dt

ZZ
+ (x, t, y(x, t), u(x, t)) ds dt, (7)
P

4
Subject to

yt y + d(x, t, y) = v in Q
X
v y + b(x, t, y) = u on (8)

y(0) = y0 in

and

va (x, t) v(x, t) vb (x, t) for a.e (x, t) Q (9)

and keeping u = u fixed,then v obeys necessary conditions for the distributed control problem

with variable v. On the other hand, if v is kept fixed, then u satisfies the necessary conditions

for the corresponding problem with boundary control u [7].

1.1.5 Construction of optimal control solutions

Validation of numerical methods for the solution of optimal control problems, control problems,

whose solution are known explicitly must be constructed. The constructed optimal control prob-

lems can be solved numerically where problem conditions are set by

a Bang-bang control, where the all the control functions values lie on the boundary of the

admissible set.

a Distributed control and Neumann boundary condition.

1.1.6 Convergence of the solutions

If X, k.k be a normed space, and xn inf ty

strongly convergent if there exist x X| limn kxn xk = 0.

n=1 U converges weakly to some

u U if

5
 lim = f (u)f U , (10)
n

which can also be represented by un u as n .

1.2 Preliminary

Having given the basic definition of the major elements that will form key part of the paper, now

let us define the function that will be used to achieve the set objectives of the paper.

Consider,

.(k(x)(x)) + (x)(x) = q(x) in ,

(11)
k(x)v(x).(x) + (x) = 0 on .

1.3 Objectives

The objectives of the paper are:

1. To develop the forward simulation system in our live-cell image analysis software package

Fluocell.

2. To build an interface between Fluocell and Dr. Banks OPTPDE solver software package

PLTMG.

3. To develop and characterize OPT-PDE solvers simulations generated in Fluocell.

4. To pre-process live-cell images in Fluocell and use PLTMG to reconstruct diffusion maps

for these images.

6
1.4 Research Question

How does the choice of regularizing functions affect the optimality of spatio-temporal solutions

for the diffusion equation?

7
Chapter 2

LITERATURE REVIEW

2.1 Overview

8
Chapter 3

Methodology

Fluorescent diffuse optical tomography an imaging technique where an organic specimen is lit up

by near infrared light discharged from point-like light sources (supposed optodes). For example,

optical fiber or lasers and the fluorescent marker has been presented which specifically ties to

considerations to be recognized, e.g., tumor cells. The light then diffuses through the tissue while

being scattered and consumed by inhomogeneities including the markers. The photons consumed

by the last are remitted at a different wavelength and are then transported back to the surface,

where they are caught and used to recreate a tomographic picture of the stamped tissue. Be that

as it may, the diffusive way of the photon transport makes this undertaking challenging. The

recreation would be encouraged if the photon thickness of the lighting up

Light can be made homogeneous inside a district of intrigue so that any variety conversely

must be because of the marker circulation. A comparable issue happens in photo-dynamic disease

treatment, where rather than a fluorescent marker, a photograph activable cytotoxin is utilized

to crush growth cells accurately, and homogeneous enlightenment is essential to stay away from

nearby under or overdoses.

Because of the mind-boggling surface state of natural examples, the configuration of opcodes

required to accomplish this is a long way from self-evident. Already distributed strategies de-

pended on a discrete. The approach, where a (substantial) arrangement of conceivable areas was

specified previously, from which the best areas are picked with the end goal that a particular

9
execution measure is limited; this Ssums to a combinatorial issue with exponential complex-

ity.e Comparing ideal source sizes would then be figured in a moment step. Conversely, the

measure space perfect control approach as depicted chapter 4 yields both area and magnitude of

the point sources in a single stride, without requiring an underlying attainable configuration or

specification of the coveted number of optodes.

10
Chapter 4

Results and Discussion

4.1 Equation

Consider equation,

.(k(x)(x)) + (x)(x) = q(x) in ,

(12)
k(x)v(x).(x) + (x) = 0 on .

If the domain of the equation is defined by n and H 1 . The boundary of the model

equation, i.e. has an outword normal vector given by v, with absorption coefficient a and

reduced scattering co-efficient s while the diffusion coefficient k = n[(a + s )]1 . models the

reflection of the cells at the boundary resulting from mismatch with q being the light emissions.

Setting the objective function to:

 
minqM (wc ) 12
2
k|w0 z kL2 (w0 ) + kqkM (wc ) + :0 (q). (13)

If the solution exist, unique and bounded, then solution to equation (13) subject to equation (12)

for > 0, gives

11
 q + min(0, p + ) = 0 (14)

where p is adjoint state, with subdifferential z. After discretization, the optimality condition

can thus be written as:

Ah h qh = 0




M0 h + ATh ph = M0 z, (15)




qh + min(0, , ph |w + = 0


c

2
with Ah being the stiffness matrix and M0 is bounded mass matrix in hei , ej iL (w0 ). Control q

can be removed by Newton method. Taking (k+1 , pk+1 ), in the system yields:

k+1 k
Ah Dk d
= (16)
M0 ATh pk+1 M0 z

12
REFERENCES

[1] Capoulade, J., Wachsmuth, M., Hufnagel, L., and Knop, M. Quantitative fluores-

cence imaging of protein diffusion and interaction in living cells. Nature biotechnology 29, 9

(2011), 835839.

[2] De Los Reyes, J. C., and Griesse, R. State-constrained optimal control of the three-

dimensional stationary navierstokes equations. Journal of Mathematical Analysis and Ap-

plications 343, 1 (2008), 257272.

[3] Illaste, A., Laasmaa, M., Peterson, P., and Vendelin, M. Analysis of molecular

movement reveals latticelike obstructions to diffusion in heart muscle cells. Biophysical journal

102, 4 (2012), 739748.

[4] Liberzon, D. Calculus of variations and optimal control theory: a concise introduction.

Princeton University Press, 2012.

[5] Lu, S., Ouyang, M., Seong, J., Zhang, J., Chien, S., and Wang, Y. The spatiotempo-

ral pattern of src activation at lipid rafts revealed by diffusion-corrected fret imaging. PLoS

Comput Biol 4, 7 (2008), e1000127.

[6] Remsing, C. C. Optimal control and hamilton-poisson formalism. Int. J. Pure Appl. Math

59, 1 (2010), 1117.

[7] Trltzsch, F. Optimal control of partial differential equations. Graduate studies in math-

ematics 112 (2010).

13
[8] Vendelin, M., and Birkedal, R. Anisotropic diffusion of fluorescently labeled atp in rat

cardiomyocytes determined by raster image correlation spectroscopy. American Journal of

Physiology-Cell Physiology 295, 5 (2008), C1302C1315.

14