SlideShare a Scribd company logo
1 of 25
Accelerating a System’s Biology Kernel Using
FPGAs
Muhammad Awais
11029341
Namal college Mianwali
Supervisor:
Dr. Waqar Nabi ( University of Glasgow)
Dr. Safee Ullah ( LUMS )
Motivation
 A number of computational Approach has been proposed for Modeling and Studying
Biological System.
 With the increase in the size of network of Genes, the complexity of Biological Model
increases rapidly.
 Field Programmable Gate Array (FPGAs) is one of the best to analyze and study the
behavior of Gene’s Regulatory Network due to its highly Parallel Architecture.
 In this project a Complex Model of Gene’s Regulatory network is implemented using
Verilog(HDL).
Contents
 Background
 Gene Regulatory Network of Cortical Area
 Implementation in Verilog (HDL)
 Results
 Conclusion and Future Work
 Question & Answer
Cerebral Cortex
 Divided into many Functionally Distinct Areas Characterized by Different combination of
genes expression.
 Genetic mechanisms plays an important role in the development of these area.
 My focus will be on the earliest stage of Arealisation: How the patterns of Gene
Expression form early in cortical Area Development.
Computational modelling of gene regulatory networks
 Why Computational Modelling ?
 Complex Behavior is difficult to understand
 Simple to implement and to Use
 To systematically screen many possible networks.
 To predict which regulatory interactions between these genes are important.
 Illuminates the design principles of the gene network regulating cortical area development.
Boolean Network Model Approach
 Boolean Variable:
Representing Genes and Proteins can take only two values i.e. (0 ,1)
 Boolean Function
 It determines a Boolean-valued output based on certain logical operations.
 The basic logical operations include AND, OR and NOT. Operators e.g
D=(A OR B) AND NOT C
 Consists of a set of Boolean Variables
{σ1, σ2, σ3, σ4 . . . . . . . , σn}
value of each variable is determined by other variable through a set of Boolean function .
F = {f1, f2, f3, f4, . . . . . . . , fn }
B is the Boolean function corresponding to variable
 One function is assigned to a one variable
Boolean Network Model Dynamics
 Boolean Network is a Graph Consisting of G( V, B)
• Node represents transcription factor
• Edges Represents regulatory input
• Boolean Gate represents Genes expression
 X( t+1) = ( A or C, A and C, not(A) or B)
 By giving an initial conditions to variable, it reaches to the stable state
where
Xi (t)= Xi ( t+1)
010 001 101 110
Boolean Network Model Dynamics
Trajectory
• Series of State Vector Transition
 Fixed Point Attractor: A single state that repeats itself
 Limit Cycle Attractor: the system visits the same finite set of states
periodically
010 001 101 110
111
010 001 101 110
Boolean Network Model of Genes Regulatory
Network for Cortical Area Development
 Development Occurs in Two-Dimensional Field
 Experiments focus on anterior-posterior patterning
 Along the anterior-posterior axis, gradients of Fgf8, Emx2, Pax6, Coup-tfi, and Sp8 play a
particularly strong role in specifying areal identity
 Fgf8, Emx2 , Pax6, Coup_tfi , sp8 : Expressed in Gradient across the
Surface of the Cortex.
Proposed Design Methodology Steps
Interactions of
Genes
Logical Rule
Hardware Description
(DHL)
Verification
Xilinx
Simulation
FPGA
Interactions of Genes
 Genes of interests
• Fgf8, Emx2 , Pax6, Coup_tfi , sp8
 Combination of interaction Between 5 Genes
• 25 = 32
 Some interaction were not considered such as
Emx2  Pax6 or Emx2 ----| Pax6
 24 interactions are assumed
+ve : inductive integrations
-ve : repressive interactions
Text in italic : Genes ( Fgf8, )
Text in up right : Proteins (Fgf8)
Possible Interactions
 According to the table , 24 Possible interactions are summarized
  represents the Inductive interaction
 ---|Represents the Repressive interaction
 24 Possible interactions form 224 (1.68*107) networks
Logical Rules or Transformed Boolean Function
 24 Possible interactions will be converted to set of Boolean Logical
Function using logical operator
 ---| ( repressive interaction Deals with Not Operator)
 Multiple regulator are combined thorough AND Operator
 A protein only be active If it corresponding gene is active at previous time
step.
 A gene is active when its transcriptional activator is active.
 Eg .. (Fgf8 Fgf8, Emx2---| Fgf8, Sp8  Fgf8, and Coup-tfi ---|Fgf8)
Fgf8 = Fgf8 and not(Emx2) and not(Coup-tfi) .
Initial States and Desired Steady Sates of Anterior and
Posterior
 State of the system is Represented with ten tuple of ‘1’ and ‘0’.
 The State of Network will be [ Fgf8 , Fgf8 , Exm2, Emx2, Pax6, Pax6, Coup-tfi, Coup-
tfi. Sp8, Sp8]
 Initial state of Anterior Compartment is [ 1, 1, 0, 0, 0, 0, 0, 0, 0 ,0]
 Initial state of Posterior Compartment is [ 0, 0, 0, 0, 0, 0, 0, 0, 0 ,0]
 Steady state of Anterior [ 1, 1, 0, 0, 1, 1, 0, 0, 1 ,1]
 Steady state of Posterior [ 0, 0, 1,1 ,0 , 0, 1, 1, 0 ,0]
Implementation in Verilog (HDL)
Finite State Machine
 Compute the state of the system at time t+1
 Each network is tested either it follow the trajectory from initial state to final States
or not.
Flow Chart Code in Verilog
Dynamics of Boolean Networks and analysis of its
output
 1.68*107 networks are Simulated.
 Good and Bad Network
Implementation of Networks
 Dynamic of the Regulatory Network is implemented.
 State of Gene depends on the interaction of its regulator at time t.
 Network is Converted to the Boolean Logic Function.
Results
 Initial Condition of Anterior [ 11 0 0 0 0 0 0 0 0 ] = 768
Best Performing Network
 This network is Good.
 Interaction of Gene were translated into
set of Boolean logic Function
Result:
 Initial State of Anterior [ 11 0 0 0 0 0 0 0 0 ] = 768
Result
Initial State of Posterior = [ 0 0 0 0 0 0 0 0 0 0 ] = 0
Conclusion & Future Recommendation
 Boolean networks for the Combination of Gene’s interaction are simulated.
 Out of 1.68*107 network, 50559 Networks that Follow the Trajectory from initial
states to Steady states.
 To find the Combination of interaction of Genes for Good and Bad Networks
THANK YOU

More Related Content

Similar to MAwais_presentaion

Robust Immunological Algorithms for High-Dimensional Global Optimization
Robust Immunological Algorithms for High-Dimensional Global OptimizationRobust Immunological Algorithms for High-Dimensional Global Optimization
Robust Immunological Algorithms for High-Dimensional Global Optimization
Mario Pavone
 
EEL4851writeup.doc
EEL4851writeup.docEEL4851writeup.doc
EEL4851writeup.doc
butest
 

Similar to MAwais_presentaion (20)

Optimal parameter selection for unsupervised neural network using genetic alg...
Optimal parameter selection for unsupervised neural network using genetic alg...Optimal parameter selection for unsupervised neural network using genetic alg...
Optimal parameter selection for unsupervised neural network using genetic alg...
 
Robust Immunological Algorithms for High-Dimensional Global Optimization
Robust Immunological Algorithms for High-Dimensional Global OptimizationRobust Immunological Algorithms for High-Dimensional Global Optimization
Robust Immunological Algorithms for High-Dimensional Global Optimization
 
03 Synthesis (1).ppt
03 Synthesis  (1).ppt03 Synthesis  (1).ppt
03 Synthesis (1).ppt
 
LINEAR FEEDBACK SHIFT REGISTER GENETICALLY ADJUSTED FOR SEQUENCE COPYING
LINEAR FEEDBACK SHIFT REGISTER GENETICALLY ADJUSTED FOR SEQUENCE COPYINGLINEAR FEEDBACK SHIFT REGISTER GENETICALLY ADJUSTED FOR SEQUENCE COPYING
LINEAR FEEDBACK SHIFT REGISTER GENETICALLY ADJUSTED FOR SEQUENCE COPYING
 
Linear Feedback Shift Register Genetically Adjusted for Sequence Copying
Linear Feedback Shift Register Genetically Adjusted for Sequence CopyingLinear Feedback Shift Register Genetically Adjusted for Sequence Copying
Linear Feedback Shift Register Genetically Adjusted for Sequence Copying
 
LSTM_public.pdf
LSTM_public.pdfLSTM_public.pdf
LSTM_public.pdf
 
On The Application of Hyperbolic Activation Function in Computing the Acceler...
On The Application of Hyperbolic Activation Function in Computing the Acceler...On The Application of Hyperbolic Activation Function in Computing the Acceler...
On The Application of Hyperbolic Activation Function in Computing the Acceler...
 
Fuzzy Logic Controller for Modern Power Systems
Fuzzy Logic Controller for Modern Power SystemsFuzzy Logic Controller for Modern Power Systems
Fuzzy Logic Controller for Modern Power Systems
 
A temporal classifier system using spiking neural networks
A temporal classifier system using spiking neural networksA temporal classifier system using spiking neural networks
A temporal classifier system using spiking neural networks
 
EEL4851writeup.doc
EEL4851writeup.docEEL4851writeup.doc
EEL4851writeup.doc
 
A011130109
A011130109A011130109
A011130109
 
Fuzzy logic
Fuzzy logicFuzzy logic
Fuzzy logic
 
AI Lesson 32
AI Lesson 32AI Lesson 32
AI Lesson 32
 
Lesson 32
Lesson 32Lesson 32
Lesson 32
 
Optimization of Fuzzy Logic controller for Luo Converter using Genetic Algor...
Optimization of Fuzzy Logic controller for Luo Converter using  Genetic Algor...Optimization of Fuzzy Logic controller for Luo Converter using  Genetic Algor...
Optimization of Fuzzy Logic controller for Luo Converter using Genetic Algor...
 
B010411016
B010411016B010411016
B010411016
 
tsopze2011
tsopze2011tsopze2011
tsopze2011
 
02 s r agents
02 s r agents02 s r agents
02 s r agents
 
Digital Cells
Digital CellsDigital Cells
Digital Cells
 
Genetic Algorithms
Genetic AlgorithmsGenetic Algorithms
Genetic Algorithms
 

MAwais_presentaion

  • 1. Accelerating a System’s Biology Kernel Using FPGAs Muhammad Awais 11029341 Namal college Mianwali Supervisor: Dr. Waqar Nabi ( University of Glasgow) Dr. Safee Ullah ( LUMS )
  • 2. Motivation  A number of computational Approach has been proposed for Modeling and Studying Biological System.  With the increase in the size of network of Genes, the complexity of Biological Model increases rapidly.  Field Programmable Gate Array (FPGAs) is one of the best to analyze and study the behavior of Gene’s Regulatory Network due to its highly Parallel Architecture.  In this project a Complex Model of Gene’s Regulatory network is implemented using Verilog(HDL).
  • 3. Contents  Background  Gene Regulatory Network of Cortical Area  Implementation in Verilog (HDL)  Results  Conclusion and Future Work  Question & Answer
  • 4. Cerebral Cortex  Divided into many Functionally Distinct Areas Characterized by Different combination of genes expression.  Genetic mechanisms plays an important role in the development of these area.  My focus will be on the earliest stage of Arealisation: How the patterns of Gene Expression form early in cortical Area Development.
  • 5. Computational modelling of gene regulatory networks  Why Computational Modelling ?  Complex Behavior is difficult to understand  Simple to implement and to Use  To systematically screen many possible networks.  To predict which regulatory interactions between these genes are important.  Illuminates the design principles of the gene network regulating cortical area development.
  • 6. Boolean Network Model Approach  Boolean Variable: Representing Genes and Proteins can take only two values i.e. (0 ,1)  Boolean Function  It determines a Boolean-valued output based on certain logical operations.  The basic logical operations include AND, OR and NOT. Operators e.g D=(A OR B) AND NOT C  Consists of a set of Boolean Variables {σ1, σ2, σ3, σ4 . . . . . . . , σn} value of each variable is determined by other variable through a set of Boolean function . F = {f1, f2, f3, f4, . . . . . . . , fn } B is the Boolean function corresponding to variable  One function is assigned to a one variable
  • 7. Boolean Network Model Dynamics  Boolean Network is a Graph Consisting of G( V, B) • Node represents transcription factor • Edges Represents regulatory input • Boolean Gate represents Genes expression  X( t+1) = ( A or C, A and C, not(A) or B)  By giving an initial conditions to variable, it reaches to the stable state where Xi (t)= Xi ( t+1) 010 001 101 110
  • 8. Boolean Network Model Dynamics Trajectory • Series of State Vector Transition  Fixed Point Attractor: A single state that repeats itself  Limit Cycle Attractor: the system visits the same finite set of states periodically 010 001 101 110 111 010 001 101 110
  • 9. Boolean Network Model of Genes Regulatory Network for Cortical Area Development  Development Occurs in Two-Dimensional Field  Experiments focus on anterior-posterior patterning  Along the anterior-posterior axis, gradients of Fgf8, Emx2, Pax6, Coup-tfi, and Sp8 play a particularly strong role in specifying areal identity  Fgf8, Emx2 , Pax6, Coup_tfi , sp8 : Expressed in Gradient across the Surface of the Cortex.
  • 10. Proposed Design Methodology Steps Interactions of Genes Logical Rule Hardware Description (DHL) Verification Xilinx Simulation FPGA
  • 11. Interactions of Genes  Genes of interests • Fgf8, Emx2 , Pax6, Coup_tfi , sp8  Combination of interaction Between 5 Genes • 25 = 32  Some interaction were not considered such as Emx2  Pax6 or Emx2 ----| Pax6  24 interactions are assumed +ve : inductive integrations -ve : repressive interactions Text in italic : Genes ( Fgf8, ) Text in up right : Proteins (Fgf8)
  • 12. Possible Interactions  According to the table , 24 Possible interactions are summarized   represents the Inductive interaction  ---|Represents the Repressive interaction  24 Possible interactions form 224 (1.68*107) networks
  • 13. Logical Rules or Transformed Boolean Function  24 Possible interactions will be converted to set of Boolean Logical Function using logical operator  ---| ( repressive interaction Deals with Not Operator)  Multiple regulator are combined thorough AND Operator  A protein only be active If it corresponding gene is active at previous time step.  A gene is active when its transcriptional activator is active.  Eg .. (Fgf8 Fgf8, Emx2---| Fgf8, Sp8  Fgf8, and Coup-tfi ---|Fgf8) Fgf8 = Fgf8 and not(Emx2) and not(Coup-tfi) .
  • 14. Initial States and Desired Steady Sates of Anterior and Posterior  State of the system is Represented with ten tuple of ‘1’ and ‘0’.  The State of Network will be [ Fgf8 , Fgf8 , Exm2, Emx2, Pax6, Pax6, Coup-tfi, Coup- tfi. Sp8, Sp8]  Initial state of Anterior Compartment is [ 1, 1, 0, 0, 0, 0, 0, 0, 0 ,0]  Initial state of Posterior Compartment is [ 0, 0, 0, 0, 0, 0, 0, 0, 0 ,0]  Steady state of Anterior [ 1, 1, 0, 0, 1, 1, 0, 0, 1 ,1]  Steady state of Posterior [ 0, 0, 1,1 ,0 , 0, 1, 1, 0 ,0]
  • 16. Finite State Machine  Compute the state of the system at time t+1  Each network is tested either it follow the trajectory from initial state to final States or not.
  • 17. Flow Chart Code in Verilog
  • 18. Dynamics of Boolean Networks and analysis of its output  1.68*107 networks are Simulated.  Good and Bad Network
  • 19. Implementation of Networks  Dynamic of the Regulatory Network is implemented.  State of Gene depends on the interaction of its regulator at time t.  Network is Converted to the Boolean Logic Function.
  • 20. Results  Initial Condition of Anterior [ 11 0 0 0 0 0 0 0 0 ] = 768
  • 21. Best Performing Network  This network is Good.  Interaction of Gene were translated into set of Boolean logic Function
  • 22. Result:  Initial State of Anterior [ 11 0 0 0 0 0 0 0 0 ] = 768
  • 23. Result Initial State of Posterior = [ 0 0 0 0 0 0 0 0 0 0 ] = 0
  • 24. Conclusion & Future Recommendation  Boolean networks for the Combination of Gene’s interaction are simulated.  Out of 1.68*107 network, 50559 Networks that Follow the Trajectory from initial states to Steady states.  To find the Combination of interaction of Genes for Good and Bad Networks

Editor's Notes

  1. Fda;lfk;lsfosd]=