Fig 1.
Elements of a communication system and signal detection model.
A digital communication system consisting of source, transmitter, channel, receiver, and sink is used to study a cellular communication system consisting of surface receptors, signaling pathways, transcription factors, and target gene expression. The communication system can be modeled as a signal detection problem consisting of signal generation and transmission through a channel before detection. Figure created with Biorender.
Table 1.
Summary of data. Flow cytometry gating strategy provided in S1 Text.
Fig 2.
Illustration of experiment, measurement, modeling, and analysis process.
Peripheral blood mononuclear cells from healthy donors and ER+ breast cancer patients were subjected to stimulation with one of 5 cytokines for 15 minutes. Cell surface markers and intracellular proteins were analyzed with spectral flow cytometry to establish cell identity and to measure response to cytokine stimulation. The response to any one of the cytokines is modeled as a sum of signal and noise. All stimulations and responses are combined to create a 6-dimensional signal response space, illustrated here in 3 dimensions. The probability of error, or signal misidentification (), and signal-to-noise ratio (SNR) are computed from the signal response space for each sample and compared across cell types, healthy donors, and breast cancer patient samples. Figure created with Biorender.
Fig 3.
Probability of error and signal-to-noise ratio are altered in peripheral blood immune cells in ER+ breast cancer patients as compared to healthy donors.
A. Signal detection is characterized by and SNR for naïve CD4 + , CD8 + , T cells, naïve B cells, NK cells, and classical monocytes in peripheral blood samples from 19 ER+ breast cancer patients and 32 healthy donors. Each datapoint corresponds to one healthy donor (grey) or ER+ breast cancer (blue) sample, integrating 6 different phosphorylation events from each of the 5 cytokine stimulations and baseline (no stimulation). B. Pairwise comparisons of the negative log transformed probability of error (
) and SNR for HD and BC for each cell subtype (unpaired t-tests adjusted for multiple comparisons,
). For illustration in bar graphs, samples with
have
.
Fig 4.
Patterns of signaling error rates and SNR across immune cells in ER+ breast cancer.
A. Immune cells cluster in and SNR by type and can be grouped into low and high signal and error combinations. B. Hierarchical clustering of signaling error across immune cell subtypes reveals patterns of error rates (columns) and increased error rates across immune cell subtypes (rows) that distinguish breast cancer samples from healthy donors. Colorbar indicates row-wise z-score.
Fig 5.
JAK1/2 inhibition induces error rates and reduces SNR in healthy donors to levels comparable to breast cancer.
A. Immune cell subtypes for ER+ breast cancer samples (blue), healthy donors (grey) and a subset of 10 healthy donor samples also treated with ruxolitinib (red). Signal detection error rates are increased, and signal fidelity is decreased in all cell subtypes except for the error rate in naïve CD4 + T cells. B. Plotting the 10 samples treated with ruxolitinib on the same SNR- axes reveal relative error and signal fidelity characteristics. C. Comparisons of SNR and error rates for all cell subtypes in BC, HD, and HD + rux, and D. comparison of HD and HD + rux (unpaired t-tests adjusted for multiple comparisons,
). For illustration in the bar graphs, samples with
have
.