Characterization and sorting for particle analyzers
Non-parametric transforms such as t-distributed stochastic neighbor embedding (tSNE) are used to analyze multi-parametric data such as data derived from flow cytometry or other particle analysis systems and methods. These transforms may be included for dimensionality reduction and identification of subpopulations (e.g., gating). By nature, non-parametric transforms cannot transform new observations without training a new transformation based on the entire dataset including the new observations. The features described parameterize non-parametric transforms using a neural network thereby allowing a small training dataset to be transformed using non-parametric techniques. The training dataset may then be used to generate an accurate parametric model for assessing additional events in a manner consistent with the initial events.
1. A method comprising:
introducing a sample into a flow cytometer;
flowing the introduced sample in a flow stream;
irradiating the sample in the flow stream with a light source;
detecting light from particles in the sample flowing in the flow stream;
collecting a first raw data set based on light detected from particles in the sample flowing in the flow stream;
applying a non-parametric transformation to the first raw data set from the flow cytometer to generate a first transformed data set;
generating a second transformed data set from the first transformed data set;
generating a transformation model using the second transformed data set; and
collecting a second raw data set based on light detected from particles in the sample flowing in the flow stream;
applying the transformation model to the second raw data set from the flow cytometer;
classifying particles based at least in part on results of applying the transformation model; and
configuring the flow cytometer to sort particles based at least in part on results of particle classification.
2. The method according to claim 1 , wherein the non-parametric transformation comprises t-distributed stochastic neighbor embedding (tSNE).
3. The method according to claim 1 , wherein the first raw data set comprises data from 10,000 detected cells or less from the flow cytometer.
4. The method according to claim 1 , wherein generating the second transformed data set comprises adding a noise component.
5. The method according to claim 4 , wherein generating the second transformed data set comprises adding the noise component to one or more of the first raw data set, the first transformed data set and a combination thereof.
6. The method according to claim 5 , wherein generating the second transformed data set comprises adding the noise component to a data pair that comprises a raw data point and transformed data point.
7. The method according to claim 6 , wherein the raw data point and the transformed data point are selected by random selection.
8. The method according to claim 7 , wherein the raw data point and the transformed data point are selected by a weighted random selection.
9. The method according to claim 8 , wherein the random selection is weighted by one or more of density and cell population.
10. The method according to claim 4 , wherein the noise component is based on a probability distribution.
11. The method according to claim 10 , wherein the probability distribution is selected from the group consisting of uniform, Gaussian and Poisson.
12. The method according to claim 1 , wherein the transformation model is a dynamic algorithm.
13. The method according to claim 12 , wherein the dynamic algorithm is a machine learning algorithm.
14. The method according to claim 1 , wherein the transformation model is applied to data from the flow cytometer in real time.
15. The method according to claim 1 , wherein the method is performed by an integrated circuit device selected from the group consisting of a field programmable gate array (FPGA), an application specific integrated circuit (ASIC) and a complex programmable logic device (CPLD).
16. The method according to claim 1 , wherein the non-parametric transformation comprises a dimensionality reduction algorithm.
17. The method according to claim 1 , wherein the transformation model comprises a dimensionality reduction model.
18. A system comprising:
one or more processing devices; and
a computer-readable storage medium comprising instructions that, when executed by the one or more processing devices, causes the system to,
receive, from a particle analyzer, measurements for a first portion of particles associated with an experiment;
convert the measurements using a non-parametric transformation into initial transformed measurements;
generate a parametric model to receive as an input raw measurements for the first portion of particles associated with the experiment and generate as an output the initial transformed measurements;
configure the particle analyzer to:
convert a measurement for a particle included in a second portion of the particles based at least in part on the parametric model;
generate a control signal to adjust an operational state of an analytical means included in the particle analyzer; and
transmit the control signal to the analytical means to achieve the operational state.
19. The system according to claim 18 , wherein the analytical means comprises sorting electronics.
20. The system according to claim 18 , wherein the analytical means comprises a fluidics system.
21. An integrated circuit programmed to:
collect a first raw data set based on light detected from irradiating particles in a sample flowing in a flow stream of a flow cytometer;
apply a non-parametric transformation to the first raw data set from the flow cytometer to generate a first transformed data set;
generate a second transformed data set from the first transformed data set;
generate a transformation model using the second transformed data set;
collect a second raw data set based on light detected from irradiating particles in the sample flowing in the flow stream of the flow cytometer;
apply the transformation model to the second raw data set from the flow cytometer;
classify particles based at least in part on results of applying the transformation model;
generating a sort decision to sort particles based at least in part on results of particle classification; and
transmitting the sort decision to the flow cytometer.