Lymph Node Metastasis

We also built gene expression reference profiles from tumor-infiltrating cells. These are based on the single-cell RNA-Seq data from Tirosh and colleagues (Tirosh et al., 2016 (link)) described above. We only used the non-lymphoid tissue samples to build these tumor-infiltrating cell’s profiles, avoiding in this way potential ‘normal immune cells’ present in the lymph nodes and spleen. These reference profiles (Supplementary file 2) were built in the same way as described above for the reference profiles of circulating immune cells, but based on the mean and standard deviation instead of median and interquartile range respectively, due to the nature of single-cell RNA-Seq data and gene dropout present with such technique.
When testing EPIC with these profiles for the single-cell RNA-Seq datasets, for the samples of primary tumor and other non-lymph node metastases, a leave-one-out procedure was applied: for each donor we built reference cell profiles based only on the data coming from the other donors.

An optimal set of predictors was chosen using a leave-one-out cross-validation procedure performed on the training set (Additional file 1). Class prediction using k genes was carried out using a diagonal linear discriminant analysis method [25 (link)], which is a variant of the standard maximum-likelihood discrimination rule. The class predictor score S is computed from the top k genes. A patient with S > 0 is assigned to the poor-prognosis group, and otherwise to the good prognosis group. We will thus refer to S as the poor-prognostic score or the risk score.
To investigate whether the risk score had an independent predictive value over the standard clinical variables, the risk score S (high–low, with 'high' defined as S > 0) was included in a multivariate logistic regression analysis with 5-year status as the outcome variable. To obtain unbiased estimates, the scores for patients in the training set were computed from the leave-one-out procedure; because of dependence between samples, however, this procedure tends to produce optimistic standard errors [26 ]. We did not attempt to correct the standard errors, because the result was also validated in independent datasets. The clinical variables were the age at diagnosis, tumor grade, tumor size and lymph node metastasis, estrogen receptor status (positive–negative) and progesterone receptor status (positive–negative). The tumor size and lymph node metastases were entered into the model in the form of a stage variable. These clinical predictors were initially compared between the good-prognosis and poor-prognosis groups.
Unsupervised hierarchical clustering of the training data was used to identify flexible risk groups; here we used the Euclidean distance with complete linkage. For validation data, we used supervised clustering based on the assignment of samples to the cluster with the closest centroid. The standard Euclidean distance was used for Uppsala datasets, but for the van't Veer dataset, because of different scales and possible outliers, the distance was based on Spearman rank correlation.
To obtain a better description of the prognosis of the patients during the follow-up, we also performed survival analysis, enabling us to use full survival information not just the 5-year status. The Cox proportional hazard model was used to assess the additional contribution of the prognosis score after adjusting for the clinical variables.

Detailed baseline and clinicopathological information, including sex, age, tumor location, tumor size, pathological type, differentiation, lymph node metastasis, and TNM stage of the patients with pancreatic diseases and HC, were obtained from the medical records of the inpatients or outpatients. The preoperative hematological parameters and liver function tests included neutrophils (× 10⁹/L), lymphocytes (× 10⁹/L), monocytes (× 10⁹/L), platelets (× 10⁹/L), plasma fibrinogens (g/L), serum albumins (g/L), prealbumin (mg/L), and CA199 (U/L) within seven days before surgery (average 2—7 days) were gathered from the medical records. TNM staging was performed using the 8th edition of the AJCC Cancer Staging Manual for Pancreatic Cancer.

Medical data collected for each patient in order to correlate biological data (including PDTO response to treatment) with clinical response are indicated in Table 2.Table 2

Medical data collected for each patient

Clinical parameters	Age, sex, weight, height, ECOG performance status, Exposure to tobacco and alcohol, Cancer location and TNM-8 classification
Histological diagnosis	Type, squamous differentiation, keratinization type, p16 status
Biologic parameters	Serum neutrophils, albumin and leukocytes levels
Pathological pronostic factors	Surgical margin, lymph node metastasis with or without capsular rupture, vascular or lymphatic embolization, perineural invasion
Oncologic treatments	Radiotherapy (volume, dose, fraction) Chemotherapy (molecules, administration regimen)
Response to treatment	Disease-free survival, Progression-free survival, Overall survival.

Clinical information of patients, including age, body mass index, TNM stage, tumor location, pathological differentiation, lymph node metastasis, nerve invasion, vascular invasion, RAS status, and BRAF status, was retrospectively collected from medical records. Follow-up was carried out via telephone or by returning to the hospital for examination. The last follow-up was on July 12, 2021. Any metastasis with an interval of more than 3 months between the diagnosis of primary tumor (PT) and ovarian metastasis was defined as metachronous; otherwise, it was considered synchronous metastasis. The time from patients receiving first-line antitumor treatment to death from any cause was overall survival (OS), whereas the time of tumor progression, death from any cause, or time to receiving second-line treatment was progression-free survival (PFS). The Ethics Committee of Guangxi Medical University Cancer Hospital approved our study (LW2021078).