Pure life letter ideas on JAMA Oncology!

JAMA Oncology | The relationship between the lumen and basal subtypes of metastatic prostate cancer and the prognosis.
The lumen and basal types of primary prostate cancer have molecular specificity and have important clinical value in predicting treatment response, but about the two types Subtypes have not been reported in detail in metastatic castration resistant prostate cancer (mCRPC)
.

Through a retrospective study of 634 mCRPC samples, it was found that the clinical response and molecular characteristics of patients with luminal and basal mCRPC were related, especially after treatment with androgen signaling inhibitors (ASIs)
.

This study suggests that the pathological and molecular characteristics of biopsy samples from mCRPC patients can be used to indicate the prognosis and the potential effects of ASIs treatment
.

Background knowledge: metastatic castration-resistant prostate cancer (mCRPC): castration-resistant metastatic prostate cancer
.

androgen-signaling inhibitors (ASIs): Androgen signal transduction inhibitor method idea: The author uses the data of 4 mCRPC cohorts to evaluate the cavity type and the basal type.
Downloaded 3 cohorts from cBioPortal: Fred Hutchinson Cancer Research Center (FHCRC) (n = 157) [Editor's introduction: cBioPortal provides such a network resource: exploring, visualizing and analyzing multi-dimensional cancer genome data
.

Querying the interactive interface and integrating user data allows researchers to interactively explore genetic changes in different samples, genes, and pathways, and can also be linked to clinical results
.

This website also provides a graphical summary of the genetic level
.

Multi-platform provides network visualization analysis, survival analysis, and software programming entrance
.

The intuitive website interface makes complex cancer genomes convenient for researchers and clinicians who have not studied bioinformatics
.

】The cohort of neuroendocrine prostate cancer patients comes from Weill Cornell Medicine (WCM) (n = 49)
.

The mCRPC cohort is from East Coast Dream Team (ECDT) (n = 266).
The author also obtained clinical and processed RNA sequencing, mutation and copy number data from the expansion of the West Coast Dream Team (WCDT) (n=162) cohort
.

The author first converted the gene expression level into the gene ranking of each sample to standardize the genes in the cohort
.

However, a large number of effects still exist (eFigure 1A in the Supplement)
.

Therefore, the author applies the empirical Bayesian framework to batch correction, which can successfully eliminate the batch effect (eFigure1B in the Supplement)
.

Because of the lack of luminal A subtype in metastatic prostate cancer, the standard centroid-based method used to determine the subtype of primary prostate cancer cannot be directly applied to mCRPC samples
.

Therefore, the author uses unsupervised hierarchical clustering of all samples as an unbiased method [Editor's introduction: hierarchical clustering: hierarchical clustering algorithm builds clusters hierarchically according to data, forming a cluster Is a tree of nodes
.

If the hierarchical decomposition is carried out from the bottom to the top, it is called agglomerated hierarchical clustering, and the hierarchical decomposition is carried out from the top to the bottom, it is called the splitting hierarchical clustering
.

Condensed hierarchical clustering first treats each object as a cluster, and then gradually merges these clusters to form a larger cluster, until all the objects are in the same cluster or meet a certain termination condition
.

Split hierarchical clustering is the opposite.
It first puts all objects in a cluster, and then gradually divides them into smaller and smaller clusters, until each object forms a cluster by itself, or reaches a certain termination condition, such as A certain desired number of clusters is reached, or the distance between two nearest clusters exceeds a certain threshold
.

The advantages of hierarchical clustering: (1) The distance and the similarity of rules are easy to define, and there are few restrictions; (2) There is no need to set the number of clusters in advance; (3) The data can be detected at different levels of granularity, and the classes can be found.
Hierarchical relationship
.

When to use hierarchical clustering: (1) Because the hierarchical clustering algorithm requires a lot of time and space, so hierarchical is suitable for clustering of small data sets
.

(2) Hierarchical clustering can detect the data at different levels of granularity when the data set is not clear and clustered into several categories, and it is hoped that the hierarchical relationship between the categories can be discovered
.

】Using the gene expression of the typical 50 genes that make up the PAM50 signature to identify metastatic prostate cancer subtypes, breast cancer, prostate cancer, and other cancers have been examined
.

DNA changes and clinical data were not used for clustering
.

DNA changes and clinical data were not used for clustering
.

Hierarchical clustering adopts nonsquared Euclidean distance method and ward.
D2 clustering method to realize Ward clustering criterion
.

Based on the increase and decrease in expression of luminal and basal markers, the two identified clusters were identified as luminal and basal, and these categories were used for subsequent analysis
.

Then use GSEA to evaluate the way of enrichment in intraluminal samples and basement membrane samples
.

Results: [Editor's analysis: Articles related to biometrics analysis generally have some fixed thoughts and behavior patterns.
The author has collected the required data and processed it
.

The first step is to determine the lumen and basal subtypes of mCRPC.
Here, the author uses hierarchical clustering to determine the lumen and basal subtypes
.

But how to prove that this is correct? So the authors used t-SNE to compress high-dimensional data into a 2-dimensional space to prove the effectiveness of the algorithm
.

The effectiveness of the algorithm is verified by visual intuition
.

It also derives the idea and content of result one
.

Result 1: Lumen and basal subtypes The authors collected and analyzed RNA sequencing data of a total of 634 mCRPC samples in 4 cohorts
.

Figure 1A depicts the expression of the PAM50 gene in all samples
.

The hierarchical clustering of the samples identified two different clusters: a basal group with 346 samples (55%), indicating higher expression of basal markers (PTTG1, CDC20, ORC6L, KIF2C, UBE2C, MELK, BIRC5, NUF2, CEP55, EXO1, CENPF, NDC80, TYMS, UBE2T, ANLN, CCNB1, RRM2, and MKI67), 288 cases (45%) in the luminal group had low basic gene expression, and the expression of luminal markers was high to mixed (ESR1 , PGR, BCL2, FOXA1, CDC6, CXXC5, MLPH, MAPT, NAT1, MDM2, MMP11, and BLVRA)
.

Gene set enrichment analysis of luminal and basal prostate cancer genes in the literature further confirmed that luminal genes were enriched in luminal subtypes (GSEA, P<< span="">0.
001)
.

Basal type genes are enriched in basal subtypes
.

(GSEA,P=0.
001)
.

The authors' results showed that only two distinct clusters in metastatic prostate cancer were consistent with the low metastatic tendency of luminal A tumors, as opposed to the three clusters found in primary prostate cancer
.

Next, the authors examined samples classified as small cell/neuroendocrine prostate cancer (SCNC) (96 samples lack these data)
.

Among 59 cases of SCNC tumors, 53 cases (90%) were basal tumors
.

Then, the author used the t-distributed stochastic neighbor-embedding (TSNE) graph, which reduces the dimensionality of RNA sequencing data so that samples with similar RNA expression profiles are close to each other in the two-dimensional graph
.

The tSNE plot in Figure 1B determines that no matter which unsupervised method is used, the two main sample clusters—basic samples and luminal samples—confirm that the authors’ hierarchical clustering results are similar
.

However, the distribution along the lumen-basal axis appears to be more continuous, rather than two distinct subpopulations
.

SCNC tumors are the most basic tumors along this axis
.

Most SCNC tumors classified as lumen appear near the junction of lumen and basal
.

These results suggest that there are luminal subtypes and basic subtypes of mCRPC
.

Figure 1.
Luminal and Basal Subtypes of Metastatic Castration-Resistant Prostate Cancer (mCRPC) [Editor's Analysis: Molecular characteristics are an important analysis method for cancer biorecognition, and they also have the most important clinical significance
.

Results Two authors evaluated the molecular differences between the two groups based on the pathways rich in RNA sequencing data, as well as the mutations and copy number changes of key prostate cancer oncogenes and tumor suppressor genes
.

The authors used GSEA to analyze the pathways and concluded that the androgen response pathway is the most significant pathway for luminal enrichment.
In addition, the major changes in key prostate cancer oncogenes and tumor suppressor genes are based on such huge data.
Reliability is self-evident
.

】The molecular characteristics of luminal subtype and basal type GSEA (Figure 2A and 2B) confirm that the androgen response pathway is the most significant way for luminal type enrichment, which is consistent with the previously published primary prostate cancer data and basal cell carcinoma SCNC
.

Figure 2.
Gene Set Enrichment Analysis The most significant pathway of basal enrichment is related to cell cycle and cell division, which is consistent with the increased expression of proliferation genes (such as mki67)
.

Then, the authors detected the somatic mutations and copy numbers of the three most frequently mutated or changed copy number tumor suppressor genes (PTEN, RB1, and tp53) and oncogenes (AR, FOXA1, and myc) in prostate cancer (Figure 3)
.

Figure 3.
The author of DNA Alterations and Subtypes found a significantly higher rate of 2rb1 changes in basal samples
.

PTEN and TP53 are missing in the two types, and there is no significant difference
.

When examining the major prostate cancer oncogenes, the authors found that foxa1 has a significantly higher rate of change in basal tumors
.

Compared with myc, AR has no significant difference
.

[Editor's Analysis: Survival difference is closely related to clinical practice, and it is also the link between life information analysis and clinical practice
.

At present, most articles will use Cox proportional hazards model or KM survival curve for survival analysis
.

Here the author uses the Cox proportional hazard model to analyze patients who have received androgen signaling inhibitor (ASI) treatment, and compares the survival between the luminal subtype and the basal subtype
.

] Survival differences between luminal subtypes and basal subtypes.
Next, the author examined the overall survival, subtypes, and treatment after the first biopsy of 80 ECDT patients and 123 WCDT patients
.

All patients examined in the ECDT cohort were protaxin and received first-line ASI treatment after biopsy
.

In the Cox proportional hazard analysis, the overall survival of patients with luminal tumors in ECDT was better than that of patients with basal tumors, with a median overall survival of 33.
1 months and a median overall survival of 18.
7 months
.

In the WCDT cohort, there were patients who received ASI treatment and non-ASI treatment after biopsy
.

In ECDT, compared with basal tumors, the overall survival of ASI treatment after intraluminal tumor biopsy is better, with a median overall survival of 21.
7 months and 32.
0 months for basal tumors
.

However, after biopsy of patients who did not receive ASI treatment, the difference was not statistically significant (Figure 4B)
.

Figure 4B.
Clinical Outcomes and Subtypes In patients with intracavitary tumors, the authors found that the survival rate of patients treated with ASI was significantly better than that of patients not treated with ASI
.

In patients with basal tumors, it is no longer significant
.

It is suggested that the efficacy of ASI treatment on intraluminal tumors may be higher than that of basal tumors
.

The median overall survival of ECDT patients and WCDT patients treated with ASI after biopsy were similar
.

Therefore, the authors combined the two cohorts for Cox proportional hazard interaction analysis and adjusted the exposure to ASI treatment before biopsy
.

The authors found that consistent with true predictive biomarkers, the interaction terms between subtype and ASI treatment after biopsy were statistically significant
.

When the authors performed a subgroup analysis of the early ASI treatment cohort, among patients who had never received ASI treatment, patients with intraluminal tumors had a better overall survival rate than patients with basal tumors
.

However, when the authors examined patients previously treated with ASI, the difference between intraluminal tumors and basal tumors disappeared
.

In order to confirm that random genes will not produce the same result, we re-analyzed with 50 random genes and repeated 100 times
.

The authors found that none of these attempts could identify important p-values ​​for survival analysis and interaction terms
.

[Editor's analysis: In the data analysis downloaded by the author, there are still 633 patients with information about metastatic sites
.

The authors developed a similarity score and found significant differences in the luminal-basal score of the biopsy site (Supplementary Figure 2)
.

] Subtype variants have information about metastatic sites in 633 patients in all 4 cohorts
.

Supplementary Figure 2 In a further pairwise comparison, the lumal-basal score in the lymph node samples was significantly increased compared with the samples from the primary site, indicating a more basal-like sample
.

This result is consistent with the author's finding that the basal subtype has a stronger aggressive phenotype
.

The FHCRC cohort is a rapid autopsy cohort that analyzes multiple tumor sites in each patient
.

This unique data set allows analysis of intra-individual variation of subtypes
.

Figure 5 depicts the variability of luminal-basal scores between and within the FHCRC cohort
.

Only 7 of 42 patients (17%) had both intraluminal and basal tumors, which indicates that these subtypes are relatively stable in individual patients
.

Figure 5.
Intrapatient Subtype Variability Summary of this article: These findings represent the largest comprehensive clinical, transcriptomic and genomic analysis of mCRPC samples so far, and are important for better providing personalized treatment for mCRPC patients based on tumor molecular information One step
.

If you are interested, please consult relevant project personnel for great benefit in November, 15% off all projects.
Related consultation hotline: 010-87520456 Transcriptome | Methylation | Resequencing | Single cell | m6A | Multiomics cytoscape | limma | WGCNA | Water Bear Bug Legend | Linux Electrophoresis | PCR | A Brief History of Sequencing | Karyotype | NIPT | Basic Experimental Genes | 2019-nCoV | Enrichment Analysis | Joint Analysis | Microenvironmental Plague Pursuit | Summary of Ideas | Scholars | Scientific Research | Withdrawal | Read Bo | Work