School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... ·...
Transcript of School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... ·...
![Page 1: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/1.jpg)
A Model for Cell Proliferation in a Developing Organism
Phil. Pollett
UQ School of Mathematics and Physics
ACEMS Research Group Meeting
19 September 2016
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 1 / 26
![Page 2: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/2.jpg)
Collaborators
Peter TaylorSchool of Mathematics and StatisticsUniversity of Melbourne
Laleh TafakoriSchool of Mathematics and StatisticsUniversity of Melbourne
Disclaimer. Work in progress!
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 2 / 26
![Page 3: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/3.jpg)
Collaborators
Peter TaylorSchool of Mathematics and StatisticsUniversity of Melbourne
Laleh TafakoriSchool of Mathematics and StatisticsUniversity of Melbourne
Disclaimer. Work in progress!
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 2 / 26
![Page 4: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/4.jpg)
A model for cell proliferation
Starting point. Talks given by Kerry Landman at recent UQ Mathematics Colloquia(August 2015 and July 2016).
The model in question is described in . . .
Hywood, J.D., Hackett-Jones, E.J. and Landman, K.A. (2013) Modeling biological tissue growth:
Discrete to continuum representations. Physical Review E , 8, 032704.
Broad aim. Improve our understanding of the role of cell proliferation in tissue growth.
Specific instance. Neural crest cells, present in the gut tissue of the developing humanembryo, form neurons that construct the enteric nervous system (ENS). Failure of thesecells to invade the gut tissue completely can cause imperfect formation of the ENS, andlead to Hirschsprung’s disease, which can result in potentially life-threatingcomplications.
Modelling. We investigate a one-dimensional lattice model for tissue growth with latticespacing ∆. (A continuum model is obtained in the limit as ∆→ 0.)
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 3 / 26
![Page 5: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/5.jpg)
A model for cell proliferation
Starting point. Talks given by Kerry Landman at recent UQ Mathematics Colloquia(August 2015 and July 2016).
The model in question is described in . . .
Hywood, J.D., Hackett-Jones, E.J. and Landman, K.A. (2013) Modeling biological tissue growth:
Discrete to continuum representations. Physical Review E , 8, 032704.
Broad aim. Improve our understanding of the role of cell proliferation in tissue growth.
Specific instance. Neural crest cells, present in the gut tissue of the developing humanembryo, form neurons that construct the enteric nervous system (ENS). Failure of thesecells to invade the gut tissue completely can cause imperfect formation of the ENS, andlead to Hirschsprung’s disease, which can result in potentially life-threatingcomplications.
Modelling. We investigate a one-dimensional lattice model for tissue growth with latticespacing ∆. (A continuum model is obtained in the limit as ∆→ 0.)
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 3 / 26
![Page 6: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/6.jpg)
A model for cell proliferation
Starting point. Talks given by Kerry Landman at recent UQ Mathematics Colloquia(August 2015 and July 2016).
The model in question is described in . . .
Hywood, J.D., Hackett-Jones, E.J. and Landman, K.A. (2013) Modeling biological tissue growth:
Discrete to continuum representations. Physical Review E , 8, 032704.
Broad aim. Improve our understanding of the role of cell proliferation in tissue growth.
Specific instance. Neural crest cells, present in the gut tissue of the developing humanembryo, form neurons that construct the enteric nervous system (ENS). Failure of thesecells to invade the gut tissue completely can cause imperfect formation of the ENS, andlead to Hirschsprung’s disease, which can result in potentially life-threatingcomplications.
Modelling. We investigate a one-dimensional lattice model for tissue growth with latticespacing ∆. (A continuum model is obtained in the limit as ∆→ 0.)
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 3 / 26
![Page 7: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/7.jpg)
A model for cell proliferation
Starting point. Talks given by Kerry Landman at recent UQ Mathematics Colloquia(August 2015 and July 2016).
The model in question is described in . . .
Hywood, J.D., Hackett-Jones, E.J. and Landman, K.A. (2013) Modeling biological tissue growth:
Discrete to continuum representations. Physical Review E , 8, 032704.
Broad aim. Improve our understanding of the role of cell proliferation in tissue growth.
Specific instance. Neural crest cells, present in the gut tissue of the developing humanembryo, form neurons that construct the enteric nervous system (ENS). Failure of thesecells to invade the gut tissue completely can cause imperfect formation of the ENS, andlead to Hirschsprung’s disease, which can result in potentially life-threatingcomplications.
Modelling. We investigate a one-dimensional lattice model for tissue growth with latticespacing ∆. (A continuum model is obtained in the limit as ∆→ 0.)
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 3 / 26
![Page 8: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/8.jpg)
A model for cell proliferation
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 4 / 26
![Page 9: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/9.jpg)
A model for cell proliferation
We investigate a one-dimensional lattice model for tissue growth with lattice spacing ∆.(A continuum model is obtained in the limit as ∆→ 0.)
Lattice site i is at position x = i∆.
Let N(t) be the number of cells at time t, so that L(t) = ∆N(t) is the length of thegrowing tissue segment.
Proliferation. A cell is selected uniformly at random. If the cell at site i is selected, itmoves to site i + 1 pushing all cells to the right of it up by one. A new unmarked cellnow occupies site i .
Times between proliferation events are iid exp(λ).
Equivalently. You can think of independent homogenous rate-λ Poisson processes ateach site triggering proliferation events.
The state X (t) of the system at time t is a binary vector of length N(t), whose i-thentry is 1 or 0 according to whether site i is occupied by a marked cell. It takes values inthe (countable) subset of {0, 1}N whose elements have only finitely many 1s.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 5 / 26
![Page 10: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/10.jpg)
A model for cell proliferation
We investigate a one-dimensional lattice model for tissue growth with lattice spacing ∆.(A continuum model is obtained in the limit as ∆→ 0.)
Lattice site i is at position x = i∆.
Let N(t) be the number of cells at time t, so that L(t) = ∆N(t) is the length of thegrowing tissue segment.
Proliferation. A cell is selected uniformly at random. If the cell at site i is selected, itmoves to site i + 1 pushing all cells to the right of it up by one. A new unmarked cellnow occupies site i .
Times between proliferation events are iid exp(λ).
Equivalently. You can think of independent homogenous rate-λ Poisson processes ateach site triggering proliferation events.
The state X (t) of the system at time t is a binary vector of length N(t), whose i-thentry is 1 or 0 according to whether site i is occupied by a marked cell. It takes values inthe (countable) subset of {0, 1}N whose elements have only finitely many 1s.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 5 / 26
![Page 11: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/11.jpg)
A model for cell proliferation
We investigate a one-dimensional lattice model for tissue growth with lattice spacing ∆.(A continuum model is obtained in the limit as ∆→ 0.)
Lattice site i is at position x = i∆.
Let N(t) be the number of cells at time t, so that L(t) = ∆N(t) is the length of thegrowing tissue segment.
Proliferation. A cell is selected uniformly at random. If the cell at site i is selected, itmoves to site i + 1 pushing all cells to the right of it up by one. A new unmarked cellnow occupies site i .
Times between proliferation events are iid exp(λ).
Equivalently. You can think of independent homogenous rate-λ Poisson processes ateach site triggering proliferation events.
The state X (t) of the system at time t is a binary vector of length N(t), whose i-thentry is 1 or 0 according to whether site i is occupied by a marked cell. It takes values inthe (countable) subset of {0, 1}N whose elements have only finitely many 1s.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 5 / 26
![Page 12: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/12.jpg)
A model for cell proliferation
We investigate a one-dimensional lattice model for tissue growth with lattice spacing ∆.(A continuum model is obtained in the limit as ∆→ 0.)
Lattice site i is at position x = i∆.
Let N(t) be the number of cells at time t, so that L(t) = ∆N(t) is the length of thegrowing tissue segment.
Proliferation. A cell is selected uniformly at random. If the cell at site i is selected, itmoves to site i + 1 pushing all cells to the right of it up by one. A new unmarked cellnow occupies site i .
Times between proliferation events are iid exp(λ).
Equivalently. You can think of independent homogenous rate-λ Poisson processes ateach site triggering proliferation events.
The state X (t) of the system at time t is a binary vector of length N(t), whose i-thentry is 1 or 0 according to whether site i is occupied by a marked cell. It takes values inthe (countable) subset of {0, 1}N whose elements have only finitely many 1s.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 5 / 26
![Page 13: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/13.jpg)
A model for cell proliferation
We investigate a one-dimensional lattice model for tissue growth with lattice spacing ∆.(A continuum model is obtained in the limit as ∆→ 0.)
Lattice site i is at position x = i∆.
Let N(t) be the number of cells at time t, so that L(t) = ∆N(t) is the length of thegrowing tissue segment.
Proliferation. A cell is selected uniformly at random. If the cell at site i is selected, itmoves to site i + 1 pushing all cells to the right of it up by one. A new unmarked cellnow occupies site i .
Times between proliferation events are iid exp(λ).
Equivalently. You can think of independent homogenous rate-λ Poisson processes ateach site triggering proliferation events.
The state X (t) of the system at time t is a binary vector of length N(t), whose i-thentry is 1 or 0 according to whether site i is occupied by a marked cell. It takes values inthe (countable) subset of {0, 1}N whose elements have only finitely many 1s.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 5 / 26
![Page 14: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/14.jpg)
A model for cell proliferation
We investigate a one-dimensional lattice model for tissue growth with lattice spacing ∆.(A continuum model is obtained in the limit as ∆→ 0.)
Lattice site i is at position x = i∆.
Let N(t) be the number of cells at time t, so that L(t) = ∆N(t) is the length of thegrowing tissue segment.
Proliferation. A cell is selected uniformly at random. If the cell at site i is selected, itmoves to site i + 1 pushing all cells to the right of it up by one. A new unmarked cellnow occupies site i .
Times between proliferation events are iid exp(λ).
Equivalently. You can think of independent homogenous rate-λ Poisson processes ateach site triggering proliferation events.
The state X (t) of the system at time t is a binary vector of length N(t), whose i-thentry is 1 or 0 according to whether site i is occupied by a marked cell. It takes values inthe (countable) subset of {0, 1}N whose elements have only finitely many 1s.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 5 / 26
![Page 15: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/15.jpg)
A model for cell proliferation
We investigate a one-dimensional lattice model for tissue growth with lattice spacing ∆.(A continuum model is obtained in the limit as ∆→ 0.)
Lattice site i is at position x = i∆.
Let N(t) be the number of cells at time t, so that L(t) = ∆N(t) is the length of thegrowing tissue segment.
Proliferation. A cell is selected uniformly at random. If the cell at site i is selected, itmoves to site i + 1 pushing all cells to the right of it up by one. A new unmarked cellnow occupies site i .
Times between proliferation events are iid exp(λ).
Equivalently. You can think of independent homogenous rate-λ Poisson processes ateach site triggering proliferation events.
The state X (t) of the system at time t is a binary vector of length N(t), whose i-thentry is 1 or 0 according to whether site i is occupied by a marked cell. It takes values inthe (countable) subset of {0, 1}N whose elements have only finitely many 1s.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 5 / 26
![Page 16: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/16.jpg)
A model for cell proliferation
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 6 / 26
![Page 17: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/17.jpg)
Q & A. What can we say about the model?
The process (X (t), t ≥ 0) is a
continuous-time Markov chain.
Consider a particular marked cell. Is its position I (t) at time t Markovian?
Yes.
If that cell is in position i , it moves to the right at rate
λi .
The position of any particular marked cell evolves as a
Yule Process, that is, apure-birth process with birth rates λi = λi (in the usual notation).
Therefore, the distance travelled by any particular marked cell up to time t (itsposition relative to its starting site j) has a
negative binomial
distribution.
That distance will be k (≥ 0) with probability
(j + k − 1
k
)e jt(1− et)
k , where et = exp(−λt).
Finally note that if we start with all marked cells in adjacent sites, then, at anyfuture time, the gaps between them will be independent, and thus the positions ofthe marked cells at any fixed time t will follow a discrete renewal process withnegative binomial lifetimes. (We have not exploited this fact in our analysis so far,but we have plans!)
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 7 / 26
![Page 18: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/18.jpg)
Q & A. What can we say about the model?
The process (X (t), t ≥ 0) is a continuous-time Markov chain.
Consider a particular marked cell. Is its position I (t) at time t Markovian?
Yes.
If that cell is in position i , it moves to the right at rate
λi .
The position of any particular marked cell evolves as a
Yule Process, that is, apure-birth process with birth rates λi = λi (in the usual notation).
Therefore, the distance travelled by any particular marked cell up to time t (itsposition relative to its starting site j) has a
negative binomial
distribution.
That distance will be k (≥ 0) with probability
(j + k − 1
k
)e jt(1− et)
k , where et = exp(−λt).
Finally note that if we start with all marked cells in adjacent sites, then, at anyfuture time, the gaps between them will be independent, and thus the positions ofthe marked cells at any fixed time t will follow a discrete renewal process withnegative binomial lifetimes. (We have not exploited this fact in our analysis so far,but we have plans!)
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 7 / 26
![Page 19: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/19.jpg)
Q & A. What can we say about the model?
The process (X (t), t ≥ 0) is a continuous-time Markov chain.
Consider a particular marked cell. Is its position I (t) at time t Markovian? Yes.
If that cell is in position i , it moves to the right at rate
λi .
The position of any particular marked cell evolves as a
Yule Process, that is, apure-birth process with birth rates λi = λi (in the usual notation).
Therefore, the distance travelled by any particular marked cell up to time t (itsposition relative to its starting site j) has a
negative binomial
distribution.
That distance will be k (≥ 0) with probability
(j + k − 1
k
)e jt(1− et)
k , where et = exp(−λt).
Finally note that if we start with all marked cells in adjacent sites, then, at anyfuture time, the gaps between them will be independent, and thus the positions ofthe marked cells at any fixed time t will follow a discrete renewal process withnegative binomial lifetimes. (We have not exploited this fact in our analysis so far,but we have plans!)
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 7 / 26
![Page 20: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/20.jpg)
Q & A. What can we say about the model?
The process (X (t), t ≥ 0) is a continuous-time Markov chain.
Consider a particular marked cell. Is its position I (t) at time t Markovian? Yes.
If that cell is in position i , it moves to the right at rate λi .
The position of any particular marked cell evolves as a
Yule Process, that is, apure-birth process with birth rates λi = λi (in the usual notation).
Therefore, the distance travelled by any particular marked cell up to time t (itsposition relative to its starting site j) has a
negative binomial
distribution.
That distance will be k (≥ 0) with probability
(j + k − 1
k
)e jt(1− et)
k , where et = exp(−λt).
Finally note that if we start with all marked cells in adjacent sites, then, at anyfuture time, the gaps between them will be independent, and thus the positions ofthe marked cells at any fixed time t will follow a discrete renewal process withnegative binomial lifetimes. (We have not exploited this fact in our analysis so far,but we have plans!)
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 7 / 26
![Page 21: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/21.jpg)
Q & A. What can we say about the model?
The process (X (t), t ≥ 0) is a continuous-time Markov chain.
Consider a particular marked cell. Is its position I (t) at time t Markovian? Yes.
If that cell is in position i , it moves to the right at rate λi .
The position of any particular marked cell evolves as a Yule Process, that is, apure-birth process with birth rates λi = λi (in the usual notation).
Therefore, the distance travelled by any particular marked cell up to time t (itsposition relative to its starting site j) has a
negative binomial
distribution.
That distance will be k (≥ 0) with probability
(j + k − 1
k
)e jt(1− et)
k , where et = exp(−λt).
Finally note that if we start with all marked cells in adjacent sites, then, at anyfuture time, the gaps between them will be independent, and thus the positions ofthe marked cells at any fixed time t will follow a discrete renewal process withnegative binomial lifetimes. (We have not exploited this fact in our analysis so far,but we have plans!)
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 7 / 26
![Page 22: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/22.jpg)
Q & A. What can we say about the model?
The process (X (t), t ≥ 0) is a continuous-time Markov chain.
Consider a particular marked cell. Is its position I (t) at time t Markovian? Yes.
If that cell is in position i , it moves to the right at rate λi .
The position of any particular marked cell evolves as a Yule Process, that is, apure-birth process with birth rates λi = λi (in the usual notation).
Therefore, the distance travelled by any particular marked cell up to time t (itsposition relative to its starting site j) has a negative binomial distribution.
That distance will be k (≥ 0) with probability
(j + k − 1
k
)e jt(1− et)
k , where et = exp(−λt).
Finally note that if we start with all marked cells in adjacent sites, then, at anyfuture time, the gaps between them will be independent, and thus the positions ofthe marked cells at any fixed time t will follow a discrete renewal process withnegative binomial lifetimes. (We have not exploited this fact in our analysis so far,but we have plans!)
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 7 / 26
![Page 23: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/23.jpg)
Q & A. What can we say about the model?
The process (X (t), t ≥ 0) is a continuous-time Markov chain.
Consider a particular marked cell. Is its position I (t) at time t Markovian? Yes.
If that cell is in position i , it moves to the right at rate λi .
The position of any particular marked cell evolves as a Yule Process, that is, apure-birth process with birth rates λi = λi (in the usual notation).
Therefore, the distance travelled by any particular marked cell up to time t (itsposition relative to its starting site j) has a negative binomial distribution.
That distance will be k (≥ 0) with probability(j + k − 1
k
)e jt(1− et)
k , where et = exp(−λt).
Finally note that if we start with all marked cells in adjacent sites, then, at anyfuture time, the gaps between them will be independent, and thus the positions ofthe marked cells at any fixed time t will follow a discrete renewal process withnegative binomial lifetimes. (We have not exploited this fact in our analysis so far,but we have plans!)
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 7 / 26
![Page 24: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/24.jpg)
Q & A. What can we say about the model?
The process (X (t), t ≥ 0) is a continuous-time Markov chain.
Consider a particular marked cell. Is its position I (t) at time t Markovian? Yes.
If that cell is in position i , it moves to the right at rate λi .
The position of any particular marked cell evolves as a Yule Process, that is, apure-birth process with birth rates λi = λi (in the usual notation).
Therefore, the distance travelled by any particular marked cell up to time t (itsposition relative to its starting site j) has a negative binomial distribution.
That distance will be k (≥ 0) with probability(j + k − 1
k
)e jt(1− et)
k , where et = exp(−λt).
Finally note that if we start with all marked cells in adjacent sites, then, at anyfuture time, the gaps between them will be independent, and thus the positions ofthe marked cells at any fixed time t will follow a discrete renewal process withnegative binomial lifetimes. (We have not exploited this fact in our analysis so far,but we have plans!)
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 7 / 26
![Page 25: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/25.jpg)
The approach of Hywood, Hackett-Jones, and Landman
They focussed attention on the expected occupancy (occupancy probability) Ci (t), thechance that site i is occupied by a marked cell at time t, and derived a continuum modelfor occupation density C(x , t) at position x (taking ∆ to its limit 0).
Simulation experiments suggested that they might approximate the movement of themarked cells by independent Yule processes, one for each marked cell .
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 8 / 26
![Page 26: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/26.jpg)
The approach of Hywood, Hackett-Jones, and Landman
They focussed attention on the expected occupancy (occupancy probability) Ci (t), thechance that site i is occupied by a marked cell at time t, and derived a continuum modelfor occupation density C(x , t) at position x (taking ∆ to its limit 0).
Simulation experiments suggested that they might approximate the movement of themarked cells by independent Yule processes, one for each marked cell .
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 8 / 26
![Page 27: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/27.jpg)
The approach of Hywood, Hackett-Jones, and Landman
They focussed attention on the expected occupancy (occupancy probability) Ci (t), thechance that site i is occupied by a marked cell at time t, and derived a continuum modelfor occupation density C(x , t) at position x (taking ∆ to its limit 0).
Simulation experiments suggested that they might approximate the movement of themarked cells by independent Yule processes, one for each marked cell .
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 8 / 26
![Page 28: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/28.jpg)
The approach of Hywood, Hackett-Jones, and Landman
Solutions, in red, to∂
∂tC(x, t) = −λ
∂
∂x[xC(x, t)] +
λ∆
2
∂2
∂x2[xC(x, t)],
and expected occupancy estimates (1000 runs), in black, for t = 1, 2, 3, 4, with
L(0) = 24, λ = 0.69, and marked cells initially in [12, 18]: (a) ∆ = 1, (b) ∆ = 1/2.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 9 / 26
![Page 29: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/29.jpg)
The expected occupancy
We observed that the expected occupancy is the same for both models, and easy to writedown.
If the marked cells are initially located at adjacent sites j = r + 1, . . . s, the expectedoccupancy for site k at time t is
Ck(t) =
min{s,k}∑j=r+1
(k − 1
k − j
)e jt(1− et)
k−j , where et = exp(−λt).
To see this, generate a Yule process starting j using an ensemble of independent rate-λPoisson processes {N(j)
i (t), i = 1, 2, . . . } (one for each site i), and superimpose theseprocesses (s − r of them).
If the {N(j)}, j = r + 1, . . . s, are independent ensembles of Poisson processes, we gettrajectories the approximating model. However, if they are the same ensemble, we gettrajectories of the original model. The above formula is a simple consequence of noticingthat the approximating model is the empirical process, which counts the numbers of“particles” in each state.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 10 / 26
![Page 30: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/30.jpg)
The expected occupancy
We observed that the expected occupancy is the same for both models, and easy to writedown.
If the marked cells are initially located at adjacent sites j = r + 1, . . . s, the expectedoccupancy for site k at time t is
Ck(t) =
min{s,k}∑j=r+1
(k − 1
k − j
)e jt(1− et)
k−j , where et = exp(−λt).
To see this, generate a Yule process starting j using an ensemble of independent rate-λPoisson processes {N(j)
i (t), i = 1, 2, . . . } (one for each site i), and superimpose theseprocesses (s − r of them).
If the {N(j)}, j = r + 1, . . . s, are independent ensembles of Poisson processes, we gettrajectories the approximating model. However, if they are the same ensemble, we gettrajectories of the original model. The above formula is a simple consequence of noticingthat the approximating model is the empirical process, which counts the numbers of“particles” in each state.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 10 / 26
![Page 31: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/31.jpg)
The expected occupancy
We observed that the expected occupancy is the same for both models, and easy to writedown.
If the marked cells are initially located at adjacent sites j = r + 1, . . . s, the expectedoccupancy for site k at time t is
Ck(t) =
min{s,k}∑j=r+1
(k − 1
k − j
)e jt(1− et)
k−j , where et = exp(−λt).
To see this, generate a Yule process starting j using an ensemble of independent rate-λPoisson processes {N(j)
i (t), i = 1, 2, . . . } (one for each site i), and superimpose theseprocesses (s − r of them).
If the {N(j)}, j = r + 1, . . . s, are independent ensembles of Poisson processes, we gettrajectories the approximating model. However, if they are the same ensemble, we gettrajectories of the original model. The above formula is a simple consequence of noticingthat the approximating model is the empirical process, which counts the numbers of“particles” in each state.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 10 / 26
![Page 32: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/32.jpg)
The expected occupancy
We observed that the expected occupancy is the same for both models, and easy to writedown.
If the marked cells are initially located at adjacent sites j = r + 1, . . . s, the expectedoccupancy for site k at time t is
Ck(t) =
min{s,k}∑j=r+1
(k − 1
k − j
)e jt(1− et)
k−j , where et = exp(−λt).
To see this, generate a Yule process starting j using an ensemble of independent rate-λPoisson processes {N(j)
i (t), i = 1, 2, . . . } (one for each site i), and superimpose theseprocesses (s − r of them).
If the {N(j)}, j = r + 1, . . . s, are independent ensembles of Poisson processes, we gettrajectories the approximating model. However, if they are the same ensemble, we gettrajectories of the original model. The above formula is a simple consequence of noticingthat the approximating model is the empirical process, which counts the numbers of“particles” in each state.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 10 / 26
![Page 33: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/33.jpg)
The expected occupancy
0 100 200 300 400 500 6000
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
0.05
Site
Occupancy
0 100 200 300 400 500 6000
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
0.05
SiteOccupancy
Estimates (blue), based on 10, 000 runs, of the expected occupancy of the pro-
liferation process (left) and the corresponding ensemble of Yule processes (right)
over 600 sites, with ∆ = 1, t = 4.0, proliferation rate λ = 0.69, and initially
7 marked cells located at sites 12 up to 18. Also plotted (solid red) is Ci (t) for
i = 1, . . . , 600.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 11 / 26
![Page 34: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/34.jpg)
The expected occupancy
0 500 1000 1500 2000 25000
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
Site
Occupancy
0 500 1000 1500 2000 25000
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
SiteOccupancy
Estimates (blue), based on 10, 000 runs, of the expected occupancy of the prolifer-
ation process (left) and the corresponding ensemble of Yule processes (right) over
2, 500 sites, with ∆ = 1, t = 4.0, proliferation rate λ = 0.69, and initially 107
marked cells located at sites 12 up to 118. Also plotted (solid red) is Ci (t) for
i = 1, . . . , 2, 500.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 12 / 26
![Page 35: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/35.jpg)
The occupancy density
Since the PDE model for the occupation density C(x , t) obtained by Hywood,Hackett-Jones, Landman was not really a continuum model, we tried to derive an explicitexpression for occupation density, as follows:
Suppose cells are situated at points in the interval [a, b] with equal spacing ∆ (> 0):
a + ∆, a + 2∆, . . . , a + N∆ (= b),
where N = (b − a)/∆ is the number of points. The idea is that the initial “cell mass”b − a is distributed evenly among these N points. We are now interested in the expectedoccupancy at position x at time t, namely
C∆(x , t) =
min{b/∆,x/∆}∑j=a/∆+1
(x/∆− 1
x/∆− j
)e jt(1− et)
x/∆−j .
What happens to C∆(x , t) as ∆→ 0?
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 13 / 26
![Page 36: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/36.jpg)
The occupancy density
Since the PDE model for the occupation density C(x , t) obtained by Hywood,Hackett-Jones, Landman was not really a continuum model, we tried to derive an explicitexpression for occupation density, as follows:
Suppose cells are situated at points in the interval [a, b] with equal spacing ∆ (> 0):
a + ∆, a + 2∆, . . . , a + N∆ (= b),
where N = (b − a)/∆ is the number of points. The idea is that the initial “cell mass”b − a is distributed evenly among these N points. We are now interested in the expectedoccupancy at position x at time t, namely
C∆(x , t) =
min{b/∆,x/∆}∑j=a/∆+1
(x/∆− 1
x/∆− j
)e jt(1− et)
x/∆−j .
What happens to C∆(x , t) as ∆→ 0?
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 13 / 26
![Page 37: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/37.jpg)
The occupancy density
Since the PDE model for the occupation density C(x , t) obtained by Hywood,Hackett-Jones, Landman was not really a continuum model, we tried to derive an explicitexpression for occupation density, as follows:
Suppose cells are situated at points in the interval [a, b] with equal spacing ∆ (> 0):
a + ∆, a + 2∆, . . . , a + N∆ (= b),
where N = (b − a)/∆ is the number of points. The idea is that the initial “cell mass”b − a is distributed evenly among these N points. We are now interested in the expectedoccupancy at position x at time t, namely
C∆(x , t) =
min{b/∆,x/∆}∑j=a/∆+1
(x/∆− 1
x/∆− j
)e jt(1− et)
x/∆−j .
What happens to C∆(x , t) as ∆→ 0?
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 13 / 26
![Page 38: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/38.jpg)
The occupancy density
0 100 200 300 400 500 6000
0.01
0.02
0.03
0.04
0.05
0.06
0.07
Position
Expectedoccupancy
0 10 20 30 40 50 60 70 80 90 100
0.04
0.045
0.05
0.055
0.06
0.065
n
Expectedoccupancy
atx=
228
The left-hand pane shows C∆(x, t) at t = 4.0 for values of ∆ from 1 down to
1/n, where n = 10. The initial cell mass is on the interval [12, 18], and λ = 0.69.
The right-hand pane shows C∆(x, t) for x = 228 (approximately where the peaks
occur) for values of ∆ = 1/n up to n = 100. The code uses nbinpdf(k-j,j,e).
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 14 / 26
![Page 39: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/39.jpg)
The occupancy density
0 500 1000 1500 2000 25000
0.01
0.02
0.03
0.04
0.05
0.06
0.07
Position
Expectedoccupancy
C∆(x, t) at t = 4.0 for values of ∆ from 1 down to 1/n, where n = 5. The initial
cell mass is on the interval [12, 118], and λ = 0.69. Decreasing ∆ corresponds to
an increasing amount of flatness in the curves and decreasing tail mass.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 15 / 26
![Page 40: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/40.jpg)
The occupancy density - an approximation
Recall that
C∆(x , t) =
min{b/∆,x/∆}∑j=a/∆+1
(x/∆− 1
x/∆− j
)e jt(1− et)
x/∆−j , where et = exp(−λt).
Suppose that all quantities are chosen so that i := x/∆− 1, l := a/∆− 1, m := b/∆− 1,and n := (b − a)/∆ = m − l are integers, and in particular when ∆ becomes small.
Then, we may write
C∆(x , t) =
min{n,i−l}∑j=1
(i
i − l − j
)θj+l+1(1− θ)i−l−j , where θ = et ,
noting that i , l , m, and n, increase at the same rate when ∆→ 0, and in particular,l/i → a/x and m/i → b/x .
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 16 / 26
![Page 41: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/41.jpg)
The occupancy density - an approximation
Recall that
C∆(x , t) =
min{b/∆,x/∆}∑j=a/∆+1
(x/∆− 1
x/∆− j
)e jt(1− et)
x/∆−j , where et = exp(−λt).
Suppose that all quantities are chosen so that i := x/∆− 1, l := a/∆− 1, m := b/∆− 1,and n := (b − a)/∆ = m − l are integers, and in particular when ∆ becomes small.
Then, we may write
C∆(x , t) =
min{n,i−l}∑j=1
(i
i − l − j
)θj+l+1(1− θ)i−l−j , where θ = et ,
noting that i , l , m, and n, increase at the same rate when ∆→ 0, and in particular,l/i → a/x and m/i → b/x .
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 16 / 26
![Page 42: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/42.jpg)
The occupancy density - an approximation
Next observe that
C∆(x , t) = θ
min{n+l,i}∑j=l+1
(i
j
)θj(1− θ)i−j = θ Pr(l + 1 ≤ Si ≤ min{n + l , i}),
where Si has a binomial B(i , θ) distribution (θ = exp(−λt)).
So, we may employ the normal approximation to the binomial distribution to approximateC∆(x , t) where ∆ is small (and hence i is large). We get C∆(x , t) ' Capprox(x , t), where
Capprox(x , t) = θ Pr
(a/x − θ√θ(1− θ)
√i ≤ Z ≤ min{b/x , 1} − θ√
θ(1− θ)
√i
),
where Z is a standard normal random variable.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 17 / 26
![Page 43: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/43.jpg)
The occupancy density - an approximation
Next observe that
C∆(x , t) = θ
min{n+l,i}∑j=l+1
(i
j
)θj(1− θ)i−j = θ Pr(l + 1 ≤ Si ≤ min{n + l , i}),
where Si has a binomial B(i , θ) distribution (θ = exp(−λt)).
So, we may employ the normal approximation to the binomial distribution to approximateC∆(x , t) where ∆ is small (and hence i is large). We get C∆(x , t) ' Capprox(x , t), where
Capprox(x , t) = θ Pr
(a/x − θ√θ(1− θ)
√i ≤ Z ≤ min{b/x , 1} − θ√
θ(1− θ)
√i
),
where Z is a standard normal random variable.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 17 / 26
![Page 44: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/44.jpg)
The occupancy density - normal approximation
0 100 200 300 400 500 600Position
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
Expectedoccupancy
0 500 1000 1500 2000 2500Position
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
Expectedoccupancy
Evaluation of C∆(x, t) at t = 4.0 with λ = 0.69, and with initial cell mass on
the interval [12, 18] (left pane) and on the interval [12, 118] (right pane). The
corresponding normal approximation is shown in bold red. The code uses normcdf.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 18 / 26
![Page 45: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/45.jpg)
The occupancy density - normal approximation
160 180 200 220 240 260 280 300 320Position
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
Expectedoccupancy
0 500 1000 1500 2000 2500Position
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
Expectedoccupancy
The normal approximation with ∆ quite small (∆ = 0.001). A clearer picture is
emerging of the shape of occupation density curve C(x, t).
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 19 / 26
![Page 46: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/46.jpg)
The occupancy density
Theorem. If, initially, the marked cells lie in the interval [a, b], the occupation density attime t is given by
C(x , t) := lim∆→0
C∆(x , t) =
0 if 0 < x < aeλt
12e−λt if x = aeλt
e−λt if aeλt < x < beλt
12e−λt if x = beλt
0 if x > beλt .
Proof . Recall that C∆(x , t) = θ Pr(l + 1 ≤ Si ≤ min{m, i}), where Si has a binomialB(i , θ) distribution (θ = exp(−λt)), and l/i → a/x and m/i → b/x as i →∞.
We use Theorem 2 of . . .
Arratia, R. and Gordon, L. (1989) Tutorial on large deviations for the binomial distribution.
Bulletin of Mathematical Biology 51, 125–131.
. . . which provides an approximation for Pr(Si ≥ ai), a > 0, when i is large.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 20 / 26
![Page 47: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/47.jpg)
The occupancy density
Theorem. If, initially, the marked cells lie in the interval [a, b], the occupation density attime t is given by
C(x , t) := lim∆→0
C∆(x , t) =
0 if 0 < x < aeλt
12e−λt if x = aeλt
e−λt if aeλt < x < beλt
12e−λt if x = beλt
0 if x > beλt .
Proof . Recall that C∆(x , t) = θ Pr(l + 1 ≤ Si ≤ min{m, i}), where Si has a binomialB(i , θ) distribution (θ = exp(−λt)), and l/i → a/x and m/i → b/x as i →∞.
We use Theorem 2 of . . .
Arratia, R. and Gordon, L. (1989) Tutorial on large deviations for the binomial distribution.
Bulletin of Mathematical Biology 51, 125–131.
. . . which provides an approximation for Pr(Si ≥ ai), a > 0, when i is large.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 20 / 26
![Page 48: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/48.jpg)
The occupancy density
Theorem. If, initially, the marked cells lie in the interval [a, b], the occupation density attime t is given by
C(x , t) := lim∆→0
C∆(x , t) =
0 if 0 < x < aeλt
12e−λt if x = aeλt
e−λt if aeλt < x < beλt
12e−λt if x = beλt
0 if x > beλt .
Proof . Recall that C∆(x , t) = θ Pr(l + 1 ≤ Si ≤ min{m, i}), where Si has a binomialB(i , θ) distribution (θ = exp(−λt)), and l/i → a/x and m/i → b/x as i →∞.
We use Theorem 2 of . . .
Arratia, R. and Gordon, L. (1989) Tutorial on large deviations for the binomial distribution.
Bulletin of Mathematical Biology 51, 125–131.
. . . which provides an approximation for Pr(Si ≥ ai), a > 0, when i is large.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 20 / 26
![Page 49: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/49.jpg)
The result of Arratia and Gordon
Let H(ε, θ) be Kullback-Leibler divergence between an ε-coin and a θ-coin:
H(ε, θ) = ε log( εθ
)+ (1− ε) log
(1− ε1− θ
).
Let r be the “odds ratio” between an ε-coin and a θ-coin:
r = r(ε, θ) =
(θ
1− θ
)/( ε
1− ε
)=θ(1− ε)ε(1− θ)
.
This satisfies 0 < r < 1 whenever θ < ε < 1.
Suppose Si has a binomial B(i , θ) distribution. If θ < ε < 1,
Pr(Si ≥ iε) ∼ 1
(1− r)√
2πε(1− ε)ie−iH(ε,θ), as i →∞.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 21 / 26
![Page 50: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/50.jpg)
The result of Arratia and Gordon
Let H(ε, θ) be Kullback-Leibler divergence between an ε-coin and a θ-coin:
H(ε, θ) = ε log( εθ
)+ (1− ε) log
(1− ε1− θ
).
Let r be the “odds ratio” between an ε-coin and a θ-coin:
r = r(ε, θ) =
(θ
1− θ
)/( ε
1− ε
)=θ(1− ε)ε(1− θ)
.
This satisfies 0 < r < 1 whenever θ < ε < 1.
Suppose Si has a binomial B(i , θ) distribution. If θ < ε < 1,
Pr(Si ≥ iε) ∼ 1
(1− r)√
2πε(1− ε)ie−iH(ε,θ), as i →∞.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 21 / 26
![Page 51: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/51.jpg)
The result of Arratia and Gordon
Let H(ε, θ) be Kullback-Leibler divergence between an ε-coin and a θ-coin:
H(ε, θ) = ε log( εθ
)+ (1− ε) log
(1− ε1− θ
).
Let r be the “odds ratio” between an ε-coin and a θ-coin:
r = r(ε, θ) =
(θ
1− θ
)/( ε
1− ε
)=θ(1− ε)ε(1− θ)
.
This satisfies 0 < r < 1 whenever θ < ε < 1.
Suppose Si has a binomial B(i , θ) distribution. If θ < ε < 1,
Pr(Si ≥ iε) ∼ 1
(1− r)√
2πε(1− ε)ie−iH(ε,θ), as i →∞.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 21 / 26
![Page 52: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/52.jpg)
The occupancy density
Theorem. If, initially, the marked cells lie in the interval [a, b], the occupation density attime t is given by
C(x , t) := lim∆→0
C∆(x , t) =
0 if 0 < x < aeλt
12e−λt if x = aeλt
e−λt if aeλt < x < beλt
12e−λt if x = beλt
0 if x > beλt .
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 22 / 26
![Page 53: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/53.jpg)
The occupancy density
0 100 200 300 400 500 6000
0.2
0.4
0.6
0.8
1
Position x
Occupancy
density
C(x,t)
Evaluation of the occupation density C(x, t) at times t = 0, 0.5, 1.0, . . . , 5.0,
with λ = 0.69, and with initial cell mass on the interval [12, 18].
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 23 / 26
![Page 54: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/54.jpg)
The occupancy density
0 100 200 300 400 500 6000
0.2
0.4
0.6
0.8
1
Position x
Occupancy
density
C(x,t)
Evaluation of the occupation density C(x, t) at times t = 0, 0.5, 1.0, . . . , 5.0,
with λ = 0.69, and with initial cell mass on the interval [12, 18]. The green bars
indicate relative cell mass.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 24 / 26
![Page 55: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/55.jpg)
The occupancy density
0 100 200 300 400 500 6000
0.2
0.4
0.6
0.8
1
Position x
Occupancy
density
C(x,t)
Evaluation of the occupation density C(x, t) at times t = 0, 0.5, 1.0, . . . , 5.0,
with λ = 0.69, and with initial cell mass on the interval [0, 18].
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 25 / 26
![Page 56: School of Mathematics and Physics - A Model for Cell Proliferation in a Developing ... · 2016-09-21 · UQ School of Mathematics and Physics ACEMS ... biological tissue growth: Discrete](https://reader034.fdocuments.net/reader034/viewer/2022050600/5fa803d32c906529ff3f252c/html5/thumbnails/56.jpg)
The occupancy density
0 100 200 300 400 500 6000
0.2
0.4
0.6
0.8
1
Position x
Occupancy
density
C(x,t)
Evaluation of the occupation density C(x, t) at times t = 0, 0.5, 1.0, . . . , 5.0,
with λ = 0.69, and with initial cell mass on the interval [0, 18]. The green bars
indicate relative cell mass.
Phil. Pollett (UQ School of Maths and Physics) A model for cell proliferation 26 / 26