Combinatorial culture condition arrays and uses thereof
AssigneeThe Broad Institute, Inc.
InventorsJesse Boehm, Yuen-Yi Tseng
Life of this patent
Estimating the term…
Abstract
Described in certain embodiments herein are combinatorial addressable arrays configured for high-throughput analysis of a sample and methods of using said combinatorial addressable arrays. Also described herein in certain embodiments are computer-implemented methods of training a statistical or machine learning model for determining and/or predicting culture conditions effective for growth of a biologic sample and computer-implemented method to determine and/or predict culture conditions effective growth for growth of a biologic sample.
Claims41
A combinatorial addressable array configured for high-throughput analysis of a sample comprising:
- an addressable array configured to receive the sample and allocate the sample to a plurality of discrete locations across the addressable array,
- wherein two or more of the discrete locations of the addressable array comprises at least two different culture conditions, and
- wherein, for each of the at least two different culture conditions, there is at least two other discrete locations on the addressable array that each comprise only that culture condition.
A computer-implemented method of training a statistical or machine learning model for determining culture conditions, predicting culture conditions, or both, effective for growth of a biologic sample, comprising:
- collecting a set of sample culture parameters from a database to generate a collected set of sample culture parameters;
- applying one or more transformations to each sample culture parameters to create a modified set of sample culture parameters;
- creating a first training set comprising the collected set of sample culture parameters, the modified set of sample culture parameters, and a set of non-effective sample culture parameter results;
- training a statistical model or a machine learning algorithm in a first stage using the first training set;
- optionally creating a second training set for a second stage of training comprising the first training set and optionally, sample culture parameters that are incorrectly detected as effective sample culture parameters after the first stage of training;
- and
- optionally training a neural network in a second stage using the second training set.
A computer-implemented method for determining culture conditions, predicting culture conditions, or both, effective for growth of a biologic sample, comprising:
- receiving biologic sample data;
- optionally applying one or more filters to the biologic sample data;
- using the received biologic sample data or filtered biologic sample data as input and applying a one or more classifiers to determine and/or predict one or more effective biologic sample biologic sample culture conditions based on a computer-accessible database, trained statistical or machine-learning model trained to predict effective biologic sample culture conditions based on the one or more classifiers, a statistical data analysis methodology, or any combination thereof.
A computer-implemented method to determine, predict, or both culture conditions effective growth for growth of a biologic sample, comprising:
- receiving data of one or more parameters from the biologic sample in a format usable by a computing device;
- executing processing logic configured to generate feature data from the received data, filter the received data, filter the feature data, process the feature data, process the received data, or any combination thereof with one or more trained machine learning models that is/are trained to predict effective biologic sample culture conditions based on the received data, feature data, or both;
- and
- executing processing logic configured to cause a list of the effective biologic sample culture conditions to be displayed via an electronic display, transmitted to a user interface program, be saved to a non-transitory computer readable memory, or any combination thereof.
Dependencies are inferred from the claim text.