Binding and Kinetics for Experimental Biologists Lecture 2 Evolutionary Computing : Initial Estimate Problem Petr Kuzmič, Ph.D. BioKin, Ltd. WATERTOWN,

Slides:



Advertisements
Similar presentations
Biologically Inspired Computing: Operators for Evolutionary Algorithms
Advertisements

CS6800 Advanced Theory of Computation
Exact and heuristics algorithms
Petr Kuzmič, Ph.D. BioKin, Ltd. WATERTOWN, MASSACHUSETTS, U.S.A. Binding and Kinetics for Experimental Biologists Lecture 1 Numerical Models for Biomolecular.
Petr Kuzmič, Ph.D. BioKin, Ltd. WATERTOWN, MASSACHUSETTS, U.S.A. Binding and Kinetics for Experimental Biologists Lecture 7 Dealing with uncertainty: Confidence.
On the Genetic Evolution of a Perfect Tic-Tac-Toe Strategy
WATERTOWN, MASSACHUSETTS, U.S.A.
Tuesday, May 14 Genetic Algorithms Handouts: Lecture Notes Question: when should there be an additional review session?
Evolutionary Computing A Practical Introduction Presented by Ben Paechter Napier University with thanks to the EvoNet Training Committee and its “Flying.
Genetic Algorithms. Some Examples of Biologically Inspired AI Neural networks Evolutionary computation (e.g., genetic algorithms) Immune-system-inspired.
A new crossover technique in Genetic Programming Janet Clegg Intelligent Systems Group Electronics Department.
Genetic Algorithms GAs are one of the most powerful and applicable search methods available GA originally developed by John Holland (1975) Inspired by.
1 Genetic Algorithms. CS The Traditional Approach Ask an expert Adapt existing designs Trial and error.
Artificial Intelligence Genetic Algorithms and Applications of Genetic Algorithms in Compilers Prasad A. Kulkarni.
Genetic Art. Network Record Using Network Information Chicago Evanston ParisMoscow.
Evolutionary Computational Intelligence
1 Genetic Algorithms. CS 561, Session 26 2 The Traditional Approach Ask an expert Adapt existing designs Trial and error.
Genetic Algorithms Nehaya Tayseer 1.Introduction What is a Genetic algorithm? A search technique used in computer science to find approximate solutions.
Intro to AI Genetic Algorithm Ruth Bergman Fall 2004.
Genetic Programming. Agenda What is Genetic Programming? Background/History. Why Genetic Programming? How Genetic Principles are Applied. Examples of.
Numerical Enzymology Generalized Treatment of Kinetics & Equilibria Petr Kuzmič, Ph.D. BioKin, Ltd. DYNAFIT SOFTWARE PACKAGE.
Parallel Genetic Algorithms with Distributed-Environment Multiple Population Scheme M.Miki T.Hiroyasu K.Hatanaka Doshisha University,Kyoto,Japan.
Genetic Algorithm.
© Negnevitsky, Pearson Education, CSC 4510 – Machine Learning Dr. Mary-Angela Papalaskari Department of Computing Sciences Villanova University.
Evolutionary Computing in the Study of Bio/Chemical Mechanisms Petr Kuzmič, Ph.D. BioKin, Ltd. Problem: Finding initial parameter estimates Solution: Differential.
WATERTOWN, MASSACHUSETTS, U.S.A.
Slides are based on Negnevitsky, Pearson Education, Lecture 12 Hybrid intelligent systems: Evolutionary neural networks and fuzzy evolutionary systems.
SOFT COMPUTING (Optimization Techniques using GA) Dr. N.Uma Maheswari Professor/CSE PSNA CET.
Improved Gene Expression Programming to Solve the Inverse Problem for Ordinary Differential Equations Kangshun Li Professor, Ph.D Professor, Ph.D College.
Intro. ANN & Fuzzy Systems Lecture 36 GENETIC ALGORITHM (1)
Lecture 8: 24/5/1435 Genetic Algorithms Lecturer/ Kawther Abas 363CS – Artificial Intelligence.
Genetic Algorithms Michael J. Watts
Genetic algorithms Charles Darwin "A man who dares to waste an hour of life has not discovered the value of life"
An Introduction to Genetic Algorithms Lecture 2 November, 2010 Ivan Garibay
Derivative Free Optimization G.Anuradha. Contents Genetic Algorithm Simulated Annealing Random search method Downhill simplex method.
EE459 I ntroduction to Artificial I ntelligence Genetic Algorithms Kasin Prakobwaitayakit Department of Electrical Engineering Chiangmai University.
© Negnevitsky, Pearson Education, Lecture 9 Evolutionary Computation: Genetic algorithms Introduction, or can evolution be intelligent? Introduction,
2005MEE Software Engineering Lecture 11 – Optimisation Techniques.
Exact and heuristics algorithms
Learning by Simulating Evolution Artificial Intelligence CSMC February 21, 2002.
 Negnevitsky, Pearson Education, Lecture 9 Evolutionary Computation: Genetic algorithms n Introduction, or can evolution be intelligent? n Simulation.
1 Genetic Algorithms K.Ganesh Introduction GAs and Simulated Annealing The Biology of Genetics The Logic of Genetic Programmes Demo Summary.
ECE 103 Engineering Programming Chapter 52 Generic Algorithm Herbert G. Mayer, PSU CS Status 6/4/2014 Initial content copied verbatim from ECE 103 material.
Genetic Algorithms. 2 Overview Introduction To Genetic Algorithms (GAs) GA Operators and Parameters Genetic Algorithms To Solve The Traveling Salesman.
Chapter 12 FUSION OF FUZZY SYSTEM AND GENETIC ALGORITHMS Chi-Yuan Yeh.
Biologically inspired algorithms BY: Andy Garrett YE Ziyu.
D Nagesh Kumar, IIScOptimization Methods: M8L5 1 Advanced Topics in Optimization Evolutionary Algorithms for Optimization and Search.
In the name of ALLAH Presented By : Mohsen Shahriari, the student of communication in Sajad institute for higher education.
1 Autonomic Computer Systems Evolutionary Computation Pascal Paysan.
An Introduction to Genetic Algorithms Lecture 2 November, 2010 Ivan Garibay
Genetic Algorithms. Underlying Concept  Charles Darwin outlined the principle of natural selection.  Natural Selection is the process by which evolution.
Genetic Programming Using Simulated Natural Selection to Automatically Write Programs.
Artificial Intelligence By Mr. Ejaz CIIT Sahiwal Evolutionary Computation.
1 Comparative Study of two Genetic Algorithms Based Task Allocation Models in Distributed Computing System Oğuzhan TAŞ 2005.
EVOLUTIONARY SYSTEMS AND GENETIC ALGORITHMS NAME: AKSHITKUMAR PATEL STUDENT ID: GRAD POSITION PAPER.
Genetic Algorithms And other approaches for similar applications Optimization Techniques.
Genetic Algorithm. Outline Motivation Genetic algorithms An illustrative example Hypothesis space search.
 Presented By: Abdul Aziz Ghazi  Roll No:  Presented to: Sir Harris.
March 1, 2016Introduction to Artificial Intelligence Lecture 11: Machine Evolution 1 Let’s look at… Machine Evolution.
Hirophysics.com The Genetic Algorithm vs. Simulated Annealing Charles Barnes PHY 327.
 Negnevitsky, Pearson Education, Lecture 12 Hybrid intelligent systems: Evolutionary neural networks and fuzzy evolutionary systems n Introduction.
Genetic (Evolutionary) Algorithms CEE 6410 David Rosenberg “Natural Selection or the Survival of the Fittest.” -- Charles Darwin.
Genetic Algorithms.
Selected Topics in CI I Genetic Programming Dr. Widodo Budiharto 2014.
Genetic Algorithms.
1 Department of Engineering, 2 Department of Mathematics,
1 Department of Engineering, 2 Department of Mathematics,
1 Department of Engineering, 2 Department of Mathematics,
GENETIC ALGORITHMS & MACHINE LEARNING
Coevolutionary Automated Software Correction
Presentation transcript:

Binding and Kinetics for Experimental Biologists Lecture 2 Evolutionary Computing : Initial Estimate Problem Petr Kuzmič, Ph.D. BioKin, Ltd. WATERTOWN, MASSACHUSETTS, U.S.A. I N N O V A T I O N L E C T U R E S (I N N O l E C)

BKEB Lec 2: Evolutionary Computing2 Lecture outline The problem: Fitting nonlinear data usually requires an initial estimate of model parameters. This initial estimate must be close enough to the “true” values. The solution: Use a data-fitting method that does not depend on initial estimates. An implementation: The Differential Evolution algorithm (Price et al., 2005). An example: Kinetics of forked DNA binding to the protein-protein complex formed by DNA-polymerase sliding clamp (gp45) and clamp loader (gp44/62).

BKEB Lec 2: Evolutionary Computing3 The ultimate goal of analyzing kinetic / binding data SELECT AMONG POSSIBLE MOLECULAR MECHANISMS concentration signal computer Select most plausible model competitive ? mechanism B mechanism C mechanism A EXPERIMENTAL DATA A VARIETY OF POSSIBLE MECHANISMS

BKEB Lec 2: Evolutionary Computing4 Most models in natural sciences are nonlinear LINEAR VS. NONLINEAR MODELS Linear y = A + k x Nonlinear y = A [1 - exp(-k x)] y = 2 [1 - exp(-5 x)] y = 2 +5 x

BKEB Lec 2: Evolutionary Computing5 We need initial estimates of model parameters NONLINEAR MODELS REQUIRE INITIAL ESTIMATES OF PARAMETERS computer k +1 k -1 k +2 k +3 k -3 A GIVEN MODEL ESTIMATED PARAMETERS k +1 k -1 k +2 k +3 k -3 REFINED PARAMETERS concentration signal EXPERIMENTAL DATA The Initial Estimate Problem: Estimated parameters most be “close enough”. How can we guess them? How can we be sure that they are “close enough”?

BKEB Lec 2: Evolutionary Computing6 The crux of the problem: Finding global minima Least-squares fitting only goes "downhill" How do we know where to start? MODEL PARAMETER SUM OF SQUARED DEVIATIONS  (data - model) 2 global minimum data - model = "residual"

BKEB Lec 2: Evolutionary Computing7 Charles Darwin to the rescue BIOLOGICAL EVOLUTION IMITATED IN "DE" ISBN-10: Charles Darwin ( )

BKEB Lec 2: Evolutionary Computing8 Specialized numerical software: DynaFit DOWNLOAD Kuzmic (2009) Meth. Enzymol., 467, DynaFit implements the Differential Evolution algorithm for global sum-of-squares minimization.

BKEB Lec 2: Evolutionary Computing9 Biological metaphor: "Gene, allele" gene BIOLOGYCOMPUTER...AAGTCG...GTAACCGG... four-letter alphabet variable length "keratin" sequence of bits representing a number two letter alphabet fixed length (16 or 32 bits) "K M ""k cat "

BKEB Lec 2: Evolutionary Computing10 "Chromosome, genotype, phenotype" genotype BIOLOGYCOMPUTER...AAGTCGGTTCGGAAGTCGGTTTA... keratin oncoprotein phenotype particular combination of all model parameters K M =4.56 V max =1.23 K is =78.9 full set of parameters

BKEB Lec 2: Evolutionary Computing11 "Organism, fitness" genotype BIOLOGYCOMPUTER...AAGTCGGTTCGGAAGTCGGTTTA... keratin oncoprotein FITNESS: agreement between the data and the model FITNESS: "agreement" with the environment

BKEB Lec 2: Evolutionary Computing12 "Population" BIOLOGYCOMPUTER low fitness V max KMKM K is medium fitness V max KMKM K is high fitness V max KMKM KisKis

BKEB Lec 2: Evolutionary Computing13 DE Population size in DynaFit number of population members per optimized model parameter number of population members per order of magnitude

BKEB Lec 2: Evolutionary Computing14 "Sexual reproduction, crossover" BIOLOGYCOMPUTER mother father "sexual mating" probability p cross child random crossover point V max KMKM K is

BKEB Lec 2: Evolutionary Computing15 "Mutation, genetic diversity" BIOLOGYCOMPUTER father V max KMKM K is mutant father V max (*) K M (*) K is (*) mutation

BKEB Lec 2: Evolutionary Computing16 "Mutation, genetic diversity" aunt #1 V max (1) K M (1) K is (1) aunt #2 V max (2) K M (2) K is (2) THE "DIFFERENTIAL" IN DIFFERENTIAL EVOLUTION ALGORITHM - STEP 1 Compute difference between two randomly chosen “auntie” phenotypes subtract aunt #2 minus aunt #1 V max (2) -V max (1) K M (2) -K M (1) K is (2) -K is (1)

BKEB Lec 2: Evolutionary Computing17 "Mutation, genetic diversity" father V max KMKM K is mutant father THE "DIFFERENTIAL" IN DIFFERENTIAL EVOLUTION ALGORITHM - STEP 2 Add weighted difference between two “uncle” phenotypes to “father” add a fraction of aunt #2 minus aunt #1 V max (2) -V max (2) K M (2) -K M (1) K is (2) -K is (1) V max * KM*KM* K is *

BKEB Lec 2: Evolutionary Computing18 "Mutation, genetic diversity" THE "DIFFERENTIAL" IN DIFFERENTIAL EVOLUTION ALGORITHM K M * = K M + F  (K M (1)  K M (2) ) EXAMPLE: Michaelis-Menten equation "father"“aunt 1"“aunt 2" "mutant father" weight (fraction) mutation rate

BKEB Lec 2: Evolutionary Computing19 DE “undocumented” settings in DynaFit probability that “child” inherits “father's” genes, not “mother's” genes These DE tuning constants are “undocumented” in the DynaFit distribution. fractional difference used in mutations K M * = K M + F  (K M (1)  K M (2) ) six different mutation strategies

BKEB Lec 2: Evolutionary Computing20 "Selection" BIOLOGYCOMPUTER high fitness more likely to breed V max KMKM KisKis more likely to be carried to the next generation low sum of squares low fitness less likely to breed V max KMKM K is less likely to be carried to the next generation high sum of squares

BKEB Lec 2: Evolutionary Computing21 Basic Differential Evolution Algorithm - Summary 1 Randomly create the initial population (size N) Repeat until almost all population members have very high fitness: 2 Evaluate fitness: sum of squares for all population members 5 Natural selection: keep child in gene pool if more fit than mother 4 Sexual reproduction: random crossover with probability P cross 3 Mutation: random gene modification (mutate father, weight F)

BKEB Lec 2: Evolutionary Computing22 Example: DNA + clamp / clamp loader complex DETERMINE ASSOCIATION AND DISSOCIATION RATE CONSTANT IN AN A + B  AB SYSTEM Courtesy of Senthil Perumal, Penn State University (Steven Benkovic lab) see Lecture 1 for details

BKEB Lec 2: Evolutionary Computing23 Example: DynaFit script for Differential Evolution INSERT A SINGLE LINE IN THE [TASK] SECTION constraints !

BKEB Lec 2: Evolutionary Computing24 Example: Initial population BOTH RATE CONSTANTS SPAN TWELVE ORDERS OF MAGNITUDE

BKEB Lec 2: Evolutionary Computing25 Example: The evolutionary process SNAPSHOTS OF k 1 / k 2 CORRELATION DIAGRAM - SPACED BY 10 “GENERATIONS”

BKEB Lec 2: Evolutionary Computing26 Example: Final population BOTH RATE CONSTANTS SPAN AT MOST ±30% RANGE RELATIVE TO NOMINAL VALUE

BKEB Lec 2: Evolutionary Computing27 Example: The “fittest” member of final population THIS IS (PRESUMABLY) THE GLOBAL MINIMUM OF SUM-OF-SQUARES data model residuals compare with “good” estimate from Lecture 1

BKEB Lec 2: Evolutionary Computing28 Example: Comparison of DE and regular data fitting DIFFERENTIAL EVOLUTION (DE) FOUND THE SAME FIT AS THE “GOOD” ESTIMATE sum of squares relative sum of sq. “best-fit” constants initial estimate k 1 = 1 k 2 = 1 k 1 = 100 k 2 = k 1 = 2.2 ± 0.5 k 2 = ± k 1 = 0.2 ± 3.4 k 2 = 0.2 ± 0.6 “good” “bad” k 1 = – k 2 = – k 1 = 2.2 ± 0.5 k 2 = ± random estimates lecture 1

BKEB Lec 2: Evolutionary Computing29 Significant disadvantage of DE: very slow DYNAFIT CAN TAKE MULTIPLE DAYS TO RUN A COMPLEX PROBLEM algorithmcomputation time Levenberg-Marquardt with two restarts Differential Evolution with four restarts (population size: 1000) 0.88 sec 12 min 31 sec DynaFit on DNA / clamp / clamp loader example: relative time 1 second 1 minute 10 minutes 15 minutes 15 hours 6 days

BKEB Lec 2: Evolutionary Computing30 Example of Differential Evolution in DynaFit J. Biol. Chem. 283, (2007) This took one week of computing on the Linux cluster in Heidelberg.

BKEB Lec 2: Evolutionary Computing31 Example: Systematic scan of many initial estimates CAREFUL! THIS IS FASTER THAN DIFFERENTIAL EVOLUTION BUT DOES NOT ALWAYS WORK 1.generate all possible combinations of rate constants 2.compute initial sum of squares for each combination 3.rank combinations by initial sum of squares 4.select the best N combinations 5.perform a full fit for those N 6.rank results again ALGORITHM 7  7 = 49 combinations of k on and k off

BKEB Lec 2: Evolutionary Computing32 Example: Systematic scan – Phase 1 AFTER EVALUATING THE INITIAL SUM OF SQUARES FOR ALL 49 COMBINATIONS OF k 1 and k 2

BKEB Lec 2: Evolutionary Computing33 Example: Systematic scan – Phase 2 AFTER RANKING THE INITIAL ESTIMATES AND SELECTING 20 BEST ONES BY SUM OF SQUARES

BKEB Lec 2: Evolutionary Computing34 Example: Systematic scan – Phase 3 AFTER PERFORMING FULL REFINEMENT FOR 20 BEST ESTIMATES OUT OF 49 TRIED “success zone” The best initial estimates do not produce the best refined solution!

BKEB Lec 2: Evolutionary Computing35 Summary and conclusions 1.Finding good-enough initial estimates is a very difficult problem. 2.One should use system-specific information as much as possible. This includes using the literature and/or general principles for “intelligent” guesses. 3.Always use the “Try” method in DynaFit to display the initial fit. Make sure that the initial estimate is at least approximately correct. 4.The Differential Evolution algorithm almost always helps. However, it can be excruciatingly slow (running typically for multiple hours). 5.The systematic scan ( task = estimate ) sometimes helps. However, the “best” initial estimates almost never produce the desired solution! 6.DynaFit is not a “silver bullet”: You must still use your brain a lot.