Granted Patent B2

US 12,609,184 · App. 17/255,722 · Granted Apr 21, 2026

Inventor: Alexander Miron (Pepper Pike, OH)

Assignee: COVARIANCE BIOSCIENCES, LLC

G16B25/20C12Q1/686G16B40/00

Abstract

Claims (138)

1 . A method of preparing an optimized primer set for multiplex genotyping, the method comprising:

A) for a given set N of variable genomic target sequences of a genome to be genotyped in a sample, designing an initial set of forward and reverse amplification primers that will amplify a sequence comprising each variable genomic target sequence in a multiplex amplification reaction, wherein the designing includes the steps of:

1) identifying all possible primers of 17 to 35 nucleotides within 100 base pairs of each genomic target sequence variation in set N of variable genomic target sequences from a pool of primers;

2) for each member of set N, selecting a subset of primer pairs from the set of step (1) that satisfies the conditions of a primer selection algorithm;

3) evaluating specificity of primer pairs chosen in step (2) in the genome, keeping only those pairs predicted to be specific for their respective targets;

4) selecting a set of optimized primers for the amplification of target gene set N, where the optimal primers are selected to minimize primer-primer interactions with other primers in the set by iterative calculation of predicted ΔG for interactions between primers to generate a fitness score and use of a fitness score optimization method selected from one or a combination of the group consisting of:

a) a Monte Carlo random or pseudo-random selection method;

b) a golden section search;

c) gradient descent;

d) minima hopping;

e) genetic algorithm;

f) neural networks;

g) cluster analysis, in which substitution is picked to minimize score; and

h) cluster analysis to create bins; and

wherein the Fitness Score is generated according to the method:

a) determining G=the set of ΔG's for all possible interactions for members of the initial primer set; and

b) calculating the Fitness Score by:

i) determining the sum, S, of |ΔG| Q for each ΔG value, wherein Q is a weighting factor constant exponent that makes large ΔG absolute values much larger than small values;

ii) determining S′=S/# of ΔG values in G;

iii) determining H=T/S′, wherein T is a constant that makes H small for large values of S′ and H large for small values of S′; and

iv) determining the Fitness Score=H R , wherein R is a weighting factor constant exponent that makes large values of H larger, and small values of H smaller; and

B) synthesizing the optimized primer set selected in step (4).

2 . The method of claim 1 , wherein steps (2)-(4) comprise:

a) for the primers identified in step (1), randomly selecting a primer pair for each target in set N that satisfies the conditions of the primer selection algorithm;

b) evaluating specificity of primer pairs chosen in step (a) in the genome, keeping only those pairs predicted to be specific for their respective targets;

c) repeating step (a) on the primer pairs kept from step (b) to generate set P, a population of randomly selected primer sets for each target in set N;

d) generating the Fitness Score for each member of population P based upon ΔG for all possible interactions between the primers in each member of the population;

e) picking member(s) of the population P based on Fitness Score;

f) repeating steps (c)-(e) iteratively until a set of primer pairs for target genes identified in step (e) has the Fitness Score at a predetermined threshold.

3 . The method of claim 1 , wherein steps (2)-(4) comprise:

a) for the primers identified in step (1), randomly selecting a primer pair for each target in set N that satisfies the conditions of the primer selection algorithm and is predicted to be specific for its target in the genome, or providing a primer pair for each target in set N, that has been selected to reduce potential for primer: primer interactions with other primers in the set and is predicted to be specific for its target in the genome;

b) repeating step (a) to generate population Z, of size 2 or greater, of primer pair sets for each target in set N;

c) generating the Fitness Score for each member of population Z based upon ΔG for all possible interactions between the primers in each member of the population;

d) selecting the members of population Z with the lowest Fitness Scores as set W;

e) replacing a primer for a single target from W with another primer identified in step (a), and generating the Fitness Score for the resulting set; wherein if the change results in an improved Fitness Score relative to the Fitness Score generated in step (c), the resulting new set W′ replaces set W, and if the change results in a no change in Fitness Score or a decreased Fitness Score, keeping set W;

f) iteratively repeating steps (c)-(e) on the set W or W′ retained in each iteration of step (e) until a set of primer pairs for target genes in set N is identified that has the Fitness Score at a predetermined threshold, or, if a predetermined threshold is not reached by iteratively repeating steps (c)-(e), beginning again at step (a) and iteratively repeating steps (c)-(e) until a set of primer pairs for target genes in set N is identified that has the Fitness Score at the predetermined threshold.

4 . The method of claim 3 , wherein the step of providing a primer pair for each target in set N that has been selected to reduce potential for primer: primer interactions with other primers in the set provides primer sets selected using one or more of a Monte Carlo random or pseudo-random selection method, a golden section search, gradient descent, minima hopping, a genetic algorithm, neural networks, cluster analysis in which substitution is picked to minimize score, or cluster analysis to create bins.

5 . The method of claim 1 , wherein steps (2)-(4) comprise:

a) generating primer set Z, including a primer pair for each member of set N either by: (i) randomly selecting from the primers identified in step (1) a primer pair for each target in set N that satisfies the conditions of the primer selection algorithm and is predicted to be specific for its target in the genome; or (ii) providing a primer pair for each target in set N that is predicted to be specific for its target in the genome, and that has been selected to reduce potential for primer: primer interactions with other primers in the set;

b) generating the Fitness Score for primer set Z based upon ΔG for all possible interactions between the primers in each member of the population;

c) making a change to a primer for a single target from set Z to generate new set Z′, and generating the Fitness Score for set Z′, wherein if the change results in an improved Fitness Score relative to that generated in step (b), the resulting new set Z′ replaces set Z, and if the change results in no change in Fitness Score or a decreased Fitness Score, keeping set Z; and

d) repeating step (c) iteratively until further iterations do not improve fitness of set Z.

6 . The method of claim 1 , wherein steps (2)-(4) comprise:

a) providing a set of optimized primer pairs for the amplification of target gene set N, where the optimal primer pairs are predicted to be specific for their target genes in the genome, and are selected to minimize primer-primer interactions with other primers in the set by iterative calculation of predicted ΔG for all possible interactions between primers to generate the Fitness Score and use of the Fitness Score optimization method selected from one or a combination of the group consisting of:

i) a Monte Carlo random or pseudo-random selection method;

ii) a golden section search;

iii) gradient descent;

iv) minima hopping;

v) genetic algorithm;

vi) neural networks;

vii) cluster analysis, in which substitution is picked to minimize score; and

viii) cluster analysis to create bins;

b) adding the set of optimized primers of step a to set M;

c) while maintaining a degree of dissimilarity from primer sets included in set M, selecting a primer pair for each target in set N from step (1) and designating it set Z, wherein the primer pairs satisfy the conditions of the primer selection algorithm, and are predicted to be specific for their target genes in the genome;

d) optimizing primer pairs of set Z for the amplification of target gene set N, to minimize primer-primer interactions with other primers in the set by iterative calculation of predicted ΔG for all possible interactions between primers to generate the Fitness Score and use of the Fitness Score optimization method selected from one or a combination of methods (i)-(viii) of step (a); and

e) repeating steps (a)-(d) iteratively until a set of primer pairs for target gene set N identified in step (d) has the Fitness Score at a predetermined threshold.

7 . The method of claim 1 , wherein steps (2)-(4) comprise:

a) for a multilayer neural network, for each primer identified in step (1) creating a node Pnz comprised by the neural network, such node connected to a node for a corresponding target (Tn), wherein

(i) each node outputs its identifier (ID) and a numeric value;

(ii) each T n produces the ID of one of the P nz nodes connected to it;

(iii) each one of the T n nodes is connected to all others; and

(iv) each node Tn is comprised by the multilayer neural network;

b) calculating the Fitness Score for output of the neural network, and on the basis of Fitness Score, the value produced by the network is compared to target, and neural network parameters for a plurality of the T n are changed;

c) calculating Fitness Score again for output of the neural network with parameters changed in step (b);

d) determining if a change was beneficial or not to the fitness of the resulting set, wherein if the change was beneficial, the direction of change is maintained with smaller increments, and wherein if the change was not beneficial, either direction is reversed or the parameters revert to a previous state;

e) repeating steps (b)-(d) iteratively, wherein at a plurality of iterations random changes are made to the parameters of the network, and wherein when the rate of fitness improvement decreases, the frequency of such random changes is increased, until a set of primer pairs for target genes in set N is identified that has the fitness score at a predetermined threshold.

8 . The method of claim 1 , wherein steps (2)-(4) comprise:

a) picking the target at random, as well as a primer for such target, and placing it in set R;

b) picking an additional target, and calculating the Fitness Score evaluating all primers for this target in combination with primers already in set R on the basis of ΔG for all potential interactions, wherein the primer that results in the best Fitness Score is added to set R;

c) if fitness of set R is below a predetermined threshold T, removing one of the primers from R according to the following:

calculating the Fitness Score for set Ri, wherein the i th target with its primer is removed from set R, and the set with the best Fitness Score determines the target with its primer to be removed from set R and placed back into the pool of primers of step (1); and

d) repeating steps (b) and (c) until all targets have optimized are assigned primers.

9 . The method of claim 1 , wherein steps (2)-(4) comprise:

a) picking the target at random, as well as a primer for such target, and placing it in set R;

b) picking an additional target, and calculating the Fitness Score evaluating all primers for this target in combination with primers already in set R on the basis of ΔG for all potential interactions, wherein the primer that results in the best Fitness Score is added to set R;

c) if fitness of set R is below a predetermined threshold T, removing one of the primers from R according to the following:

calculating the Fitness Score for set Ri, wherein the i th target with its primer is removed from set R, and the set with the best Fitness Score determines the target with its primer to be removed from set R and placed back into the pool of primers of step (1);

d) repeating steps (b) and (c) until all targets have optimized primers;

e) once all targets have optimized primers, designating set R as R 1 , and its fitness as F 1 ;

f) creating empty set R z+1 , where Z is the number of sets, with fitness F z+1 ;

g) for each set Rz, where z is an index from 1 to number of sets R, determining the element that is worst for the set's fitness, and removing this element, designated Target E;

h) recalculating Fz after removal of Target E;

i) for all Rz, determining where Target E can be added so as to maximize Fz and maximize the minimum of Fz; and

j) if the minimum of Fz is below the predetermined threshold, repeating steps (f)-(i) until the standard deviation of Fz is below the predetermined threshold, thereby designing the multiplex primer set.

10 . The method of claim 9 , wherein the step of determining the element in step (g) that is worst for fitness is performed in a method according to step (4).

11 . A method of multiplex amplification, sequencing, and/or genotyping comprising using an optimized primer set designed according to claim 2 .

12 . The method of claim 1 , wherein the optimized primer set has decreased primer-primer interactions with other primers in the set, compared to the initial set of forward and reverse amplification primers.

13 . The method of claim 1 , wherein the Fitness Score of the optimized primer set is increased compared to the initial set of forward and reverse amplification primers.

14 . The method of claim 1 , wherein the Fitness Score of the optimized primer set is a threshold Fitness Score of at least 400.

15 . The method of claim 1 , wherein the fitness score optimization method is a method of genetic algorithm,

wherein steps (2) and (3) comprise:

i) for each member of set N, selecting from the set of primers in step (1) a subset of primer pairs that satisfies the conditions for a primer selection algorithm and is predicted to be specific for its target; and

ii) repeating step (i) to generate set P, a population of randomly selected primer sets for each target gene in set N; and

wherein step (4) comprises:

iii) calculating a Fitness Score for each member of the population P; and

iv) placing members of population P into a pool of candidate primer sets on the basis of Fitness Scores; and

v) randomly selecting a plurality of “parent” sets of candidate primers from the pool of step (iv), each parent set including a different pair of candidate primer sets, parent A and parent B; and

vi) for each parent set of candidate primers, creating a crossover set of candidate primers by replacing a subset of candidate primer pairs of parent A with the corresponding subset of primer pairs of parent B; and

vii) randomly replacing one primer pair in crossover set A with a different primer pair for the corresponding target sequence generated in step (i) to create a Generation 2 population of primer sets for each target gene in set N; and

viii) repeating steps (iii)-(vii) iteratively until a set of primer pairs for target genes in set N is identified that has a Fitness Score at a predetermined threshold, and runs for an additional set amount of iterations with no measurable improvement in the fitness of the best member, whereby an optimized primer set is designed.

16 . A method of preparing an optimized primer set for multiplex genotyping, the method comprising:

A) for a given set N of variable genomic target sequences to be genotyped in a sample, designing a set of forward and reverse amplification primers that will amplify a sequence comprising each variable genomic target sequence in a multiplex amplification reaction, wherein the designing includes the steps of:

1) identifying all possible primers of 17 to 35 nucleotides within 100 base pairs of each genomic target sequence variation in set N of variable genomic target sequences;

2) For each member of set N, selecting from the set of primers in step (1) a subset of primer pairs that satisfies the conditions of a primer selection algorithm and is predicted to be specific for its target;

3) Repeating step (2) to generate set P, a population of randomly selected primer sets for each target gene in set N;

4) calculating a Fitness Score for each member of the population P;

wherein the Fitness Score is generated according to the method:

a) determining G=the set of ΔG's for all possible interactions determined for members of the primer set; and

b) calculating the Fitness Score by:

i) determining the sum, S, of |ΔG| Q for each ΔG value, wherein Q is a weighting factor constant exponent that makes large ΔG absolute values much larger than small values;

ii) determining S′=S/# of ΔG values in G;

iii) determining H=T/S′, wherein T is a constant that makes H small for large values of S′ and H large for small values of S′; and

iv) determining the Fitness Score=H R , wherein R is a weighting factor constant exponent that makes large values of H larger, and small values of H smaller;

5) Placing members of population P into a pool of candidate primer sets on the basis of Fitness Scores;

6) randomly selecting a plurality of “parent” sets of candidate primers from the pool of step (5), each parent set including a different pair of candidate primer sets, parent A and parent B;

7) For each parent set of candidate primers, creating a crossover set of candidate primers by replacing a subset of candidate primer pairs of parent A with the corresponding subset of primer pairs of parent B;

8) Randomly replacing one primer pair in crossover set A with a different primer pair for the corresponding target sequence generated in step (2) to create a Generation 2 population of primer sets for each target gene in set N; and

9) repeating steps (4)-(8) iteratively until a set of primer pairs for target genes in set N is identified that has a Fitness Score at a predetermined threshold, and runs for an additional set amount of iterations with no measurable improvement in the fitness of the best member, whereby an optimized primer set is designed; and

B) synthesizing the optimized primer set designed in step (9).

17 . A method of preparing a primer set for multiplex genotyping, the method comprising:

A) for a given set N of variable genomic target sequences to be genotyped in a sample, designing a set of forward and reverse amplification primers that will amplify a sequence comprising each variable genomic target sequence in a multiplex amplification reaction, wherein the designing includes the steps of:

1) identifying all possible primers of 17 to 35 nucleotides within 100 base pairs of each genomic target sequence variation in set N of variable genomic target sequences;

2) Selecting a primer set for the multiplex amplification and genotyping of the members of set N comprising:

a) from the set of all possible primers for each genomic target sequence variation of step (1), randomly selecting set P, a population of sets of candidate primers, each individual set of candidate primers in population P including a primer pair for the amplification of each member of set N of variable genomic target sequences to be genotyped;

b) calculating a fitness score for each member of the population of set P by calculating ΔG for all possible interactions between candidate primers in each member of the population of set P, and assigning each member of set P a Fitness Score according to the rule:

i) G=the set of ΔG's for all possible interactions determined for a given member of set P;

ii) Number of top scorers to go into next generation=1 . . . N, Number of distinct populations sets=1 . . . N, and Population size=1 . . . N such that number of top scorers to go into next generation is greater or equal to population size;

wherein the fitness score is calculated by:

iii) for each member of set P, calculating the sum, S, of |ΔG| Q for each ΔG value in that member, wherein Q is a weighting factor constant exponent that makes large ΔG absolute values much larger than small values;

iv) S′=S/# of ΔG values in G;

v) H=T/S′, wherein T is a constant that makes H small for large values of S′ and H large for small values of S′;

vi) Fitness Score=H R , wherein R is a weighting factor constant exponent that makes large values of H larger, and small values of H smaller;

c) selecting a set of primers for the multiplex amplification and genotyping of members of set N by:

i) randomly selecting a plurality of sets of “parent” sets of candidate primers, each having parent set A and parent set B, from set P based upon Fitness Scores;

ii) for each member of the plurality of sets of parents, creating a crossover set of candidate primers by replacing a subset of candidate primers in parent set A with a corresponding subset of candidate primers in parent set B, resulting in two crossover sets, crossover set A and crossover set B; and

iii) randomly replacing one primer pair in crossover set A with a different primer pair for the corresponding variable genomic target sequence to create a next generation population of candidate sets of primers, Generation 2; and

d) iteratively repeating steps (a)-(c), whereby a primer set for the multiplex amplification and genotyping of set N of variable genomic target sequences is selected; and

B) synthesizing the primer set designed in step (A).

Assignments (1)

Continuity (2)

References Cited (56)