Coding for multiplexed fluorescence microscopy
A way to design a codebook for estimating the type of a molecule at a particular location in a fluorescence microscopy image makes use of one or both of (1) knowledge of the non-uniform prior distribution of molecule types (i.e., some types are known a priori to occur more frequently than others) and/or knowledge of co-occurrence of molecule types at close locations (e.g., in a same cell); and (2) knowledge of a model of the (e.g., random) process that yields the intensities that are expected at a location when a molecule with a particular subset of markers (i.e., a molecule of a type that has been assigned a codeword that defines that subset) is present at that location. The codebook design may provide experimental efficiency by reducing the number of images that need to be acquired and/or improve classification or detection accuracy by making the codewords for different molecule types more distinctive.
1. A method for identifying molecule types comprising:
determining a marking of a set of molecule types, the marking of the set of molecule types comprising a marking of each molecule type in the set of molecule types, each marking of a molecule type comprising an assignment of a distinct set of markers from a plurality of markers;
receiving sensor data representing emissions from sample molecules in a biological sample, the sample molecules being a subset of a plurality of marked molecules, the marked molecules being marked according to the determined marking of the set of molecule types; and
identify respective molecule types of the sample molecules based on the received sensor data and marking of the set of molecule types;
wherein determining the marking of the set of molecule types comprises optimizing the markings of the set based on at least one of a prior distribution of the molecule types and co-occurrence information for the molecule types.
2. The method of claim 1 , wherein identifying respective molecule types of the sample molecules comprises applying a Bayesian decision procedure.
3. The method of claim 1 , wherein determining the marking of the set of molecule types comprises using a Reinforcement Learning (RL) procedure.
4. The method of claim 1 , wherein determining the marking of the set of molecule types comprises
forming a parameterized model of possible markings of the set of molecule types;
initializing values of parameters of the parameterized model;
repeating an iteration including (i) generating a random set of possible marking of the set of molecule types using the parameterized model, (ii) evaluating a utility of each possible marking of the set of possible markings, and (iii) updating the values of the parameters of the parameterized model; and
determining the marking of the set of molecule types from values of the parameters of the model after the repeating of the iteration.
5. The method of claim 4 , wherein generating the random set of possible marking comprises applying a Gibbs Sampling procedure or a autoregessive procedure.
6. The method of claim 4 , wherein evaluating the utility of each possible marking comprises performing a simulation of generation of sensor data from a plurality of molecule types according to the possible marking.
7. The method of claim 1 , wherein the marking of the set of molecule types is represented as a codebook, and each marking of a molecule type is represented as a codeword, the codewords together forming the codebook, and each marker of the plurality of markers is associated with a distinct position in the codewords of the codebook.
8. The method of claim 1 , wherein the markers comprise fluorescent markers, wherein each marker is responsive to at least one of a plurality of excitations, and has one of a plurality of emission characteristics.
9. The method of claim 1 , wherein the set of molecule types has a distribution characterizing a relative frequency of their presence in a sample, and wherein determining a marking of a set of molecule types includes determining said marking according to said distribution.
10. The method of claim 9 , wherein determining the marking of the set of molecule types according to the distribution comprises evaluating the utility of each possible marking using said distribution.
11. The method of claim 10 , wherein evaluating the utility of a possible marking using said distribution comprises computing a classification accuracy using the possible marking.
12. The method of claim 1 , wherein the set of molecule types has a distribution characterizing co-located occurrence of groups of two or more molecule types in a sample, and wherein determining a marking of a set of molecule types includes determining said marking according to said characterization of co-located occurrence.
13. The method of claim 12 , wherein determining the marking of the set of molecule types according to the distribution comprises evaluating the utility of each possible marking using said characterization of co-located occurrence.
14. The method of claim 13 , wherein evaluating the utility of a possible marking using said characterization of co-located occurrence comprises computing an accuracy of detection of co-location of molecule types using the possible marking.
15. The method of claim 1 , wherein:
determining the marking of the set of molecule types comprises
forming a parameterized model of possible markings of the set of molecule types;
initializing values of parameters of the parameterized model;
repeating an iteration including (i) generating a random set of possible marking of the set of molecule types using the parameterized model, (ii) evaluating a utility of each possible marking of the set of possible markings, and (iii) updating the values of the parameters of the parameterized model; and
determining the marking of the set of molecule types from values of the parameters of the model after the repeating of the iteration, including determining said marking according to a distribution characterizing co-located occurrence of groups of two or more molecule types in a sample; and
evaluating the utility of each possible marking comprises performing a simulation of generation of sensor data from a plurality of molecule types according to the possible marking; and
the marking of the set of molecule types is represented as a codebook, and each marking of a molecule type is represented as a codeword, the codewords together forming the codebook, and each marker of the plurality of markers is associated with a distinct position in the codewords of the codebook.
16. A method for determining a marking of a set of molecule types, the marking of the set of molecule types comprising a marking of each molecule type in the set of molecule types, each marking of a molecule type comprising an assignment of a distinct set of markers from a plurality of markers, the marking being determined for multiplexed fluorescence microscopy, the method comprising:
forming a parameterized model of possible markings of the set of molecule types;
initializing values of parameters of the parameterized model;
repeating an iteration including (i) generating a random set of possible marking of the set of molecule types using the parameterized model, (ii) evaluating a utility of each possible marking of the set of possible markings, and (iii) updating the values of the parameters of the parameterized model;
determining the marking of the set of molecule types from values of the parameters of the model after the repeating of the iteration; and
providing the determined marking for preparing a biological sample to be analyzed using multiplexed fluorescence microscopy.
17. A non-transitory machine-readable medium having instructions stored thereon, said instructions, when executed by a computer processor, cause the computer processor to perform steps including:
determining a marking of a set of molecule types, the marking of the set of molecule types comprising a marking of each molecule type in the set of molecule types, each marking of a molecule type comprising an assignment of a distinct set of markers from a plurality of markers;
receiving sensor data representing emissions from sample molecules in a biological sample, the sample molecules being a subset of a plurality of marked molecules, the marked molecules being marked according to the determined marking of the set of molecule types; and
identify respective molecule types of the sample molecules based on the received sensor data and marking of the set of molecule types;
wherein determining the marking of the set of molecule types comprises optimizing the markings of the set based on at least one of a prior distribution of the molecule types and co-occurrence information for the molecule types.