Close
  Indian J Med Microbiol
 

Figure 6: Pointwise mutual information maps as potential diagnostic biomarkers. Pointwise mutual information maps were constructed for individual cores using the background distributions of cell phenotypes in the entire dataset and were pooled together for patient-level pointwise mutual information (entire tumor) to better assess intratumor heterogeneity. A representative (a) estrogen receptor(+) invasive ductal carcinoma patient, (b) estrogen receptor(+) invasive lobular carcinoma patient, (c) estrogen receptor(−) invasive ductal carcinoma patient, and (d) human epidermal growth factor 2(+) invasive ductal carcinoma patient pointwise mutual information map was shown, as well as pointwise mutual information maps for the three cores taken from each patient. A heterogeneity score was assigned to each core/patient based on the entries in each pointwise mutual information map (see Methods for the relevant equation). Based on this heterogeneity score, patients AL13-3 estrogen receptor(+) invasive ductal carcinoma and AL13-6 estrogen receptor(+) invasive lobular carcinoma show more heterogeneity (difference from background distribution) than AL13-14 estrogen receptor(−) invasive ductal carcinoma and AL13-21 human epidermal growth factor 2(+) invasive ductal carcinoma. The degree to which the core-level pointwise mutual information maps change with respect to each other and the patient-level map can elucidate how much or little intratumor heterogeneity exists. For example, the core-level pointwise mutual information maps for patient AL13-14 are very similar, signifying that each core is a reasonable approximation for the patient-level analysis. As a contrary example, patient AL13-21 has highly differing core-level pointwise mutual information maps, signifying a high degree of intratumor heterogeneity in this patient. The summary heterogeneity score can provide a simple low-level understanding of heterogeneity between or within patient samples while the pointwise mutual information maps can provide a higher-level understanding, providing insight into the spatial relationships of different cell types which bring about the heterogeneity. We also propose visualization tools that can help elucidate these relationships in this higher-level understanding [Figure 5]

Figure 6: Pointwise mutual information maps as potential diagnostic biomarkers. Pointwise mutual information maps were constructed for individual cores using the background distributions of cell phenotypes in the entire dataset and were pooled together for patient-level pointwise mutual information (entire tumor) to better assess intratumor heterogeneity. A representative (a) estrogen receptor(+) invasive ductal carcinoma patient, (b) estrogen receptor(+) invasive lobular carcinoma patient, (c) estrogen receptor(−) invasive ductal carcinoma patient, and (d) human epidermal growth factor 2(+) invasive ductal carcinoma patient pointwise mutual information map was shown, as well as pointwise mutual information maps for the three cores taken from each patient. A heterogeneity score was assigned to each core/patient based on the entries in each pointwise mutual information map (see Methods for the relevant equation). Based on this heterogeneity score, patients AL13-3 estrogen receptor(+) invasive ductal carcinoma and AL13-6 estrogen receptor(+) invasive lobular carcinoma show more heterogeneity (difference from background distribution) than AL13-14 estrogen receptor(−) invasive ductal carcinoma and AL13-21 human epidermal growth factor 2(+) invasive ductal carcinoma. The degree to which the core-level pointwise mutual information maps change with respect to each other and the patient-level map can elucidate how much or little intratumor heterogeneity exists. For example, the core-level pointwise mutual information maps for patient AL13-14 are very similar, signifying that each core is a reasonable approximation for the patient-level analysis. As a contrary example, patient AL13-21 has highly differing core-level pointwise mutual information maps, signifying a high degree of intratumor heterogeneity in this patient. The summary heterogeneity score can provide a simple low-level understanding of heterogeneity between or within patient samples while the pointwise mutual information maps can provide a higher-level understanding, providing insight into the spatial relationships of different cell types which bring about the heterogeneity. We also propose visualization tools that can help elucidate these relationships in this higher-level understanding [Figure 5]