Major weakness Methodology section [1]. Methods of tissue specimen handling needs to be stated (Parafin embedded/flash freeze with liquid nitrogen or else). Transcriptomic profiling from Parafin embedded specimen can cause major loss of RNA quality. [2]. FACS - References for the CCSC marker are essential. Statistical analysis [3]. Multivariate analysis of potential factors associated with survival should be included (such as tumor stage and lymph node metastasis). Results section [4]. There is confusion in the term used in the result. For example, "To assess the quality parameters of prognostic prediction, accuracy, sensitivity, and specificity were evaluated using a univariate Logistic Regression (LR) for the 5 loci". "Prognosis" is the survival of the patients regardless of the response or non-response status. "Diagnostic" is the term for distinguishing between cancer and non-cancer. "Response/Treatment status" is the term used for separating responder and non-responder. Minor weakness Introduction section [1]. The authors can provide the definitions of non-response and response. For example, non-response (persistent/progressive disease); response (no residual tumor). [2]. Authors also can provide the supported information of CSC role in of cancer treatment and the pathways of Wnt, PI3K in both CSC and normal cancer cells. [3]. The authors should explain the unique features of CSC from normal cancer cells and the correlation between the CSC and treatment resistance. Methodology section [4]. Lack of exclusion criteria that may affect tumor profile, such as receiving prior neoadjuvant chemotherapy. Results section [5]. Figure 1 is not relevant to the objective of the study and Figure 2A is unnecessary. [6]. Supplementary data 1 should represent the difference of characteristics between "response" and "non-response" patients (dependent variable). [7]. Translating figure 3B into a volcano plot will provide more understanding of the data (please note that the caption is not English). [8]. In figure 3C, concluding the heatmap into a more interpretable figure such as Venn- diagram will help the reader to easily grasp the results. [9]. The data of AUC and pAUC in selected genes should be provided in supplementary data. [10]. Some data in text and table is not matched, maybe due to typing error. For example, "supplementary table 1" in section 3.4. Discussion section [11]. Paragraphs 1 and 2 are not necessary for discussion, but they can be integrated with the introduction. [12]. In paragraph 5, the gene expression of CSC represents the cancer dormancy. The association between the cell dormancy and CCRT resistance should be discussed. [13]. In paragraph 6, lncRNAs MALAT1, CRNDE, NEAT1, and NORAD were presented in the results as "poor prognostic indicator," but they were lowly expressed. The explanation of potential mechanisms why these genes are lowlily exhibited in the CSCs should be discussed. [14]. The discussion was mainly focused on ILF2. What about SNX2, COPZ1, ACO16722.1, AL360175.1? [15]. The main goal of the study is "treatment response." Therefore, the discussion on the prognosis context seems out of the objectives. [16]. Limitations should be stated so that they can guide other researchers to perform other research that may address these limitations. [17]. Novelty and impact section can be integrated in the last paragraph of the discussion. Finally, I would like to the authors to consider the above-mentioned major and minor weakness before acceptance the manuscript for publication. I will be happy with pleasure to review this manuscript again after the authors revision.

Waiting for submitting a second review! Regards, Kittinun Leetanaporn