US 12,241,830 B2
Living biosensors
Jesse Boehm, Cambridge, MA (US); Niklas Rindtorff, Cambridge, MA (US); JT Neal, Cambridge, MA (US); Aviad Tsherniak, Cambridge, MA (US); and Mushriq Muhib Al-Jazrawe, Cambridge, MA (US)
Assigned to Broad Institute, Inc., Cambridge, MA (US)
Filed by The Broad Institute, Inc., Cambridge, MA (US)
Filed on Dec. 7, 2020, as Appl. No. 17/113,790.
Claims priority of provisional application 62/944,880, filed on Dec. 6, 2019.
Prior Publication US 2021/0190762 A1, Jun. 24, 2021
Int. Cl. G01N 33/48 (2006.01); G01N 21/64 (2006.01); G01N 33/50 (2006.01)
CPC G01N 21/64 (2013.01) [G01N 21/6458 (2013.01); G01N 21/6486 (2013.01); G01N 33/5023 (2013.01); G01N 33/5026 (2013.01); G01N 2021/6439 (2013.01)] 20 Claims
 
1. A method of identifying cell types and/or cell phenotypes in an ex vivo sample comprising:
(a) distributing sub-samples of a cancer sample obtained from a subject into individual discrete volumes, wherein each individual discrete volume comprises a mixture of cells;
(b) obtaining one or more label-free images of the mixture of cells in each individual discrete volume;
(c) identifying cancer cells in each individual discrete volume using a morphological classifier previously trained to identify cancer cells in a sample comprising a mixture of cells;
(d) exposing each individual discrete volume to a different treatment, wherein each treatment comprises one or more drugs or a variable concentration of one or more drugs;
(e) obtaining multiple label-free images of the mixture of cells in each individual discrete volume at multiple time points, and
(f) identifying the cancer cells as sensitive or resistant to a given treatment using a convolutional neural network previously trained to extract one or more morphological features of the identified cancer cells from the obtained multiple label-free images and identify cancer cell sensitivity or resistance based on the extracted one or more morphological features, wherein identifying if the cancer cells are sensitive or resistant to the treatment comprises distinguishing between drug-induced cell death and non-drug-induced cell death.