Evaluating combination therapies in a murine BRCA1-related TNBC model.

(A.) The schematic illustrates the in vivo study design. Syngeneic olaparib-sensitive or -resistant tumors were implanted into FVB/NJ mice and treated with olaparib alone or in combination with alpelisib or Poly(I:C) once tumors reached a diameter of 5 mm. Created with BioRender.com. (B.) Kaplan-Meier curves illustrate time taken for tumors to reach study endpoint of 20 mm. i. Olaparib-sensitive tumors were treated with olaparib, alpelisib, or the combination of both. ii. Olaparib-sensitive tumors were treated with olaparib or a combination of olaparib and Poly(I:C). Similarly, olaparib-resistant tumors were treated with olaparib alone or in combination with iii. alpelisib or iv. Poly(I:C). *p<0.05, **p<0.01, ***p<0.001.

Spatial miRNA analysis overview.

(A.) Schematic of the miRNA assay used to obtain spatially-resolved miRNA data. (B.) Characteristic 7-plex spatial miRNA output. (C.) Case-study-wide LDA pipeline for assessing treatment resistance. (D.) i. LDA topic results from the case-study-wide LDA model in the form of a word cloud where the font size of each miRNA corresponds to its relative weighting for that topic and the orange font signifies a relatively greater weight within a topic as well. ii. The spatial assignment of the LDA topics to a tumor sample; the color of each well corresponds to the color of the topic borders in i. iii. The percentage of each tissue assigned to each topic. The first letter on the left corresponds to whether a tumor was known a priori to be olaparib-resistant (R) or -sensitive (S), followed by treatment group: olaparib-only (O), olaparib+alpelisib (OA), or olaparib+Poly(I:C) (OP); the color of the bars corresponds to the topic border colors. iv. PCA of all the tumor samples’ topic percentages. The samples in green correspond to tumors that were known to be sensitive to treatment, while the red denote resistant samples; the black colored text corresponds to control tumor samples. Topic 2 positively contributed towards PC1, while topic 1 contributed negatively.

Treatment-sensitive tumor analysis.

(A.) Schematic of the specialized miRNA analysis of the sensitive-only tumors. i. First, the miRNA data of olaparib-sensitive tumors are selected. ii. The miRNA data is then used to train an independent LDA model. iii. The highest miRNA probabilities from that model were extracted. iv. The extracted miRNA probabilities were then used as weights and are multiplied to the input miRNA data, thus “priming” the data. v. The case-study-wide LDA model from Fig 2C was applied to the primed miRNA data. vi. Downstream analysis was performed to stratify the olaparib-sensitive tumors. (B.) i. The spatial assignment of the case-study-wide LDA topics to the sensitive tumor samples after being primed by the specialized LDA model miRNA probabilities. The assigned topics were as in Fig 2D, where the blue corresponds to the let-7a-dominated topic (topic 1) and orange corresponds to the miR-21-dominated topic (topic 2). ii. The percentage of each tissue assigned to each topic. iii. PCA of the sensitive tumor samples’ topic percentages. The samples in green correspond to the tumors that received the combination therapies, while the red denote olaparib-only treated samples. Topic 1 was found to contribute positively to PC1, while topic 2 contributed negatively.

Spatial LDA-SSIM analysis.

(A.) Schematic of the spatial LDA-SSIM analysis pipeline. i. The pre-processed inputs of the analysis pipeline consist of the spatial miRNA data and spatial immune cell counts derived from CD45 immunostaining ii. Both the cells and spatial miRNA data were processed prior to input into the SSIM. The miRNAs were processed through the case-study-wide LDA analysis pipeline, and only the top 2 dominant spatial topics were input into SSIM (Fig 2D.ii). Cell counts were processed by calculating the ratio of immune cells to tumor cells, with appropriate normalization applied, to represent spatial immune cell infiltration within each tumor. iii. The resulting heatmap representation of the SSIM maps showing the structural similarity between the immune cell infiltration and spatial miRNA topic assignments (B.) i. LDA topic results from training an LDA model on the miRNA topic-immune infiltration SSIM maps ii. The percentage of each tissue assigned to each SSIM topic. iii. The spatial assignment of the SSIM LDA topics to the sensitive samples iv. The spatial assignment of the SSIM LDA topics to the resistant samples; the color of each well corresponds to the color of the SSIM LDA topic borders in i. while the intensity of the color corresponds to the intensity of the dominant SSIM maps that inform the topic.