TY - JOUR
T1 - Nanotechnology and machine learning enable circulating tumor cells as a reliable biomarker for radiotherapy responses of gastrointestinal cancer patients
AU - Poellmann, Michael J.
AU - Bu, Jiyoon
AU - Liu, Stanley
AU - Wang, Andrew Z.
AU - Seyedin, Steven N.
AU - Chandrasekharan, Chandrikha
AU - Hong, Heejoo
AU - Kim, Young Soo
AU - Caster, Joseph M.
AU - Hong, Seungpyo
N1 - Publisher Copyright:
© 2023 Elsevier B.V.
PY - 2023/4/15
Y1 - 2023/4/15
N2 - A highly sensitive, circulating tumor cell (CTC)-based liquid biopsy was used to monitor gastrointestinal cancer patients during treatment to determine if CTC abundance was predictive of disease recurrence. The approach used a combination of biomimetic cell rolling on recombinant E-selectin and dendrimer-mediated multivalent immunocapture at the nanoscale to purify CTCs from peripheral blood mononuclear cells. Due to the exceptionally high numbers of CTCs captured, a machine learning algorithm approach was developed to efficiently and reliably quantify abundance of immunocytochemically-labeled cells. A convolutional neural network and logistic regression model achieved 82.9% true-positive identification of CTCs with a false positive rate below 0.1% on a validation set. The approach was then used to quantify CTC abundance in peripheral blood samples from 27 subjects before, during, and following treatments. Samples drawn from the patients either prior to receiving radiotherapy or early in chemotherapy had a median 50 CTC ml−1 whole blood (range 0.6–541.6). We found that the CTC counts drawn 3 months post treatment were predictive of disease progression (p =.045). This approach to quantifying CTC abundance may be a clinically impactful in the timely determination of gastrointestinal cancer progression or response to treatment.
AB - A highly sensitive, circulating tumor cell (CTC)-based liquid biopsy was used to monitor gastrointestinal cancer patients during treatment to determine if CTC abundance was predictive of disease recurrence. The approach used a combination of biomimetic cell rolling on recombinant E-selectin and dendrimer-mediated multivalent immunocapture at the nanoscale to purify CTCs from peripheral blood mononuclear cells. Due to the exceptionally high numbers of CTCs captured, a machine learning algorithm approach was developed to efficiently and reliably quantify abundance of immunocytochemically-labeled cells. A convolutional neural network and logistic regression model achieved 82.9% true-positive identification of CTCs with a false positive rate below 0.1% on a validation set. The approach was then used to quantify CTC abundance in peripheral blood samples from 27 subjects before, during, and following treatments. Samples drawn from the patients either prior to receiving radiotherapy or early in chemotherapy had a median 50 CTC ml−1 whole blood (range 0.6–541.6). We found that the CTC counts drawn 3 months post treatment were predictive of disease progression (p =.045). This approach to quantifying CTC abundance may be a clinically impactful in the timely determination of gastrointestinal cancer progression or response to treatment.
KW - Circulating tumor cell
KW - Convolutional neural network
KW - Gastrointestinal cancer
KW - Liquid biopsy
UR - http://www.scopus.com/inward/record.url?scp=85148679172&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85148679172&partnerID=8YFLogxK
U2 - 10.1016/j.bios.2023.115117
DO - 10.1016/j.bios.2023.115117
M3 - Article
C2 - 36753988
AN - SCOPUS:85148679172
SN - 0956-5663
VL - 226
JO - Biosensors and Bioelectronics
JF - Biosensors and Bioelectronics
M1 - 115117
ER -