Back to results
US 6,640,191 B1Granted
Library design in combinatorial chemistry by Monte Carlo methods
AssigneeThe Regents of the University of California
InventorsMichael W. Deem, Marco Falcioni
Life of this patent
Estimating the term…
Abstract
Methods for generating multiple rounds of combinatorial libraries, which use Monte Carlo methods to search the multi-dimensional composition and non-composition space of variables in combinatorial chemistry; the combinatorial libraries generated and screened by such Monte Carlo methods; and an apparatus for generating and screening such libraries robotically. The process involves preparing a first set of samples, then changing the composition and non-composition variables of the samples using Monte Carlo sampling methods, and accepting proposed new samples according to a detailed balance acceptance criterion.
Claims17
A method for generating a combinatorial library, comprising:
- (a) preparing a first set of current samples, each current sample comprising a composition having two or more variables, wherein said variables are composition or non-composition variables, each variable having a value;
- (b) preparing a new set of proposed samples by changing the values of the variables of each current sample by a Monte Carlo selection method, wherein the values of the variables are randomly changed a small amount;
- (c) experimentally measuring a figure of merit for each current and proposed sample;
- and
- (d) forming a modified set of current samples according to an acceptance criterion comprising a detailed balance probability, wherein the detailed balance probability includes a Metropolis probability, whereby:
- (1) a proposed sample replaces a current sample, wherein the values of the variables of the current sample were changed to give the proposed sample, if the figure of merit of the proposed sample is greater than the figure of merit of the current sample, and
- (2) if the figure of merit of the proposed sample is less than the figure of merit of the current sample, the proposed sample is accepted according to a Metropolis probability, the Metropolis probability being determined by the formula:
- p=min{1, exp[β(Eproposed−Ecurrent)]}, (1)
- wherein β is an empirically determined parameter that defines a scale for variations of the figure of merit that are likely to be accepted, Eproposed is the figure of merit of the proposed sample and Ecurrent is the figure of merit of the current sample.
6 more elements
A method for generating a combinatorial library, comprising:
- (a) preparing a first set of current samples, each current sample comprising a composition or a molecule having two or more variables, each variable having a value;
- (b) preparing a new set of proposed samples by changing the values of the variables of the current samples by a Monte Carlo selection method, which includes at least one swapping step, wherein the values of a subset of variables is traded between two current samples, i and j, to give two proposed samples, i and j;
- (c) experimentally measuring a figure of merit for each current and proposed sample;
- and
- (d) forming a modified set of current samples according to an acceptance criterion comprising a detailed balance probability, wherein the detailed balance probability of accepting the proposed samples of the swapping step is determined by the formula:
- p=min{1, exp{β[Eiproposed+Ejproposed−Eicurrent−Ejcurrent]}, (2)
- wherein the empirically determined parameter β defines the scale of the variations of the figure of merit that are likely to be accepted, Eiproposed is the figure of merit of the proposed sample i, Ejproposed is the figure of merit of the proposed sample j, Eicurrent is the figure of merit of the current sample i and Ejcurrent is the figure of merit of the current sample j.
4 more elements
A method for generating a combinatorial library, comprising:
- (a) preparing a first set of current samples, each current sample comprising a composition having two or more variables, wherein said variables are composition or non-composition variables, each variable having a value;
- and
- (b) preparing a new set of proposed samples by changing the values of the variables of each current sample by a Monte Carlo selection method, wherein the variables are changed by:
- (1) randomly changing the value of at least one of the composition and non-composition variables of a randomly chosen current sample a small amount;
- and
- (2) swapping the values of a subset of the variables between at least one pair of randomly chosen current samples.
3 more elements
A method for generating a combinatorial library, comprising:
- (a) preparing a first set of current samples, each set of samples comprising two or more subsets, each current sample comprising a composition or a molecule having two or more variables that can be changed independently, each variable having a value;
- (b) preparing a new set of proposed samples by changing the values of the variables of each cent sample by a Monte Carlo selection method;
- (c) experimentally measuring a figure of merit for each current and proposed sample;
- (d) forming a modified set of current samples, each subset forming a modified subset of current samples, each subset being updated with a different detailed balance probability having a different empirically determined parameter β;
- which defines the scale of the variations of the figure of merit that are likely to be accepted, and
- (e) randomly exchanging two or more samples, by parallel tempering, between the subsets, wherein the exchange of two samples by parallel tempering is accepted according to an acceptance criterion derived from a detailed balance condition.
3 more elements
Dependencies are inferred from the claim text.