Patent Application
20230265522
The present invention is directed to a method for predicting the outcome of prostate carcinoma patients. The present invention (ProstaTrend) provides a method for molecular patient risk stratification after treatment, such as radical prostatectomy. The present invention also offers a method that can be used in the clinical decision-making process for potential scheduling of adjuvant therapy or evaluating biopsies in the context of active surveillance and focal therapies.
Prostate cancer is the most common solid cancer among men both in the US and in Europe. Existing diagnostic methods are often unable to precisely classify risk groups, thereby leading similarly to overdiagnosis and overtreatment as well as to insufficient detection of high-risk patients. An underlying reason is the highly variable clinical behavior of localized prostate cancer and that common classification methods of prostate cancer are often based on clinical and histological parameters only. Aggressive types of prostate cancer require radical treatment while indolent ones may be suitable for active surveillance or organ preserving focal therapies. Yet, stratification by clinical risk categories or available molecular tests lack sufficient precision. Therefore, additional molecular markers are needed to improve the stratification and outcome prediction in men with prostate cancer. Currently, tools for risk stratification based on gene expression are available and have been shown to outperform risk assessment by clinical parameters alone. An example for risk stratification tools are Oncotype (Cullen et al., 2015; Eur Urol), which predicts adverse pathology after biopsy, Decipher (Erho et al., 2013; PLoS ONE), which predicts metastasis-free survival after radical prostatectomy (RP) and Prolaris (Cuzick et al., 2015; Br J Cancer and 2011; Lancet Oncol), which is the only assay suitable for long-term prognosis after surgery, such as RP. However, none of the previously cited tools evaluate all tumor-relevant signaling pathways and therefore provide a limited molecular risk discrimination. Therefore, the problem to be solved is the persisting high clinical need for improved methods using gene expression signatures for long-term prognosis for prostate cancer patients prior and after surgery.
The present invention aims to solve the problem described above, which can be summarized as the urgent need for improved classification and risk stratification of prostate cancer patients. The solution provided by the present invention comprises the calculation of a predictive score on the basis of the expression values of up to 1396 gene transcripts (RNA biomarkers) from a set of genes associated with (prostate) cancer and (prostate) cancer patient outcome.
The inventors developed a method for calculating and applying a prognostic multi-gene expression score based on transcriptome-wide gene expression analysis (ProstaTrend), which allows improved molecular risk stratification, improved prediction of long-term prognosis and counselling of prostate cancer patients after surgery, such as radical prostatectomy, regarding adjuvant therapy options as well as evaluating biopsies prior to radical prostatectomy in the context of active surveillance and focal therapies.
The present invention provides a method (ProstaTrend) for molecular risk stratification and improved prediction of long-term prognosis for prostate cancer patients with the method comprising calculating and applying a prognostic multi-gene expression score (herein also referred to as ProstaTrend score) based on the expression of a set of genes that are significantly associated with prostate cancer prognosis in a patient sample. The calculation of the prognostic ProstaTrend score comprises the normalization and standardization of the expression values (abundancies) of marker genes (Table 2) in a patient sample and multiplication with the respective logHR values (listed in Table 2). The method disclosed by the present invention further allows the prediction of death of disease (DoD) and biochemical recurrence (BCR) for long-term prognosis and outcome for prostate cancer patients. The method provided by the invention (ProstaTrend) is independent from clinical parameters after diagnosis and facilitates the identification of high-risk patients among patients with low or intermediate histological risk profiles. The herein disclosed invention also relates to a method for use in the treatment of prostate cancer, wherein the outcome or classification of the disease of a prostate carcinoma patient from whom the sample was derived is predicted based on the calculated gene expression score value.
The invention also relates to a method for predicting the outcome or classifying the disease of a prostate carcinoma patient, the method comprising the steps of: a) obtaining a sample from said patient, wherein the sample comprises a plurality of nucleic acids, b) analyzing the plurality of nucleic acids, thereby determining the abundancy of at least one or more different target nucleic acid sequences, c) normalizing of the target nucleic acid sequence abundancies against a reference dataset and standardization, d) weighting the abundancy of the at least one or more different target nucleic acid using a corresponding reference value and e) calculating a score value, and wherein, if the score value is above a threshold value of 0, the patient is classified as having a high risk of death of disease and/or biochemical recurrence and/or if the score value is below a threshold value of 0, the patient is classified as having a low risk of death of disease and/or biochemical recurrence.
Herein, each of the one or more different target nucleic acid sequences may be identical to, without being limited to, the nucleic acid sequence of a corresponding specific reference gene, a transcript or an isoform of a reference gene, which may be selected, without being limited to, from the group of reference genes listed in Table 2.
Herein, the weighting of said abundancy of the target nucleic acids may be performed, without being limited to, by weighting said abundancy of the target nucleic acids by the logHR value of a corresponding reference gene listed in Table 2, wherein the nucleic acid sequence of said reference gene, or of a transcript or isoform thereof, is identical to the nucleic acid sequence of one or more of said target nucleic acid sequences from said sample.
Herein, a biological sample may be derived from, without being limited to, a human patient, a cell culture, an “organ-on-a-chip” model, a microphysiological system or an animal. Herein, a biological sample may be an organ, a tissue, a biopsy, a liquid biopsy, a blood sample, a sample from a Patient Derived Xenograft, a sample from an “organ-on-a-chip” model, a sample from a microphysiological system, a cell line or cell culture of a patient-derived sample of a human patient, an animal or an animal model, of a cell line or of an organoid.
Furthermore, herein the analysis of the nucleic acid sequences may comprise, but is not limited to, optional lysis of the tissue and/or the cells and optional subsequent purification of respective nucleic acids followed by optional fragmentation, enzymatic digestion or ligation of respective nucleic acids, reverse transcription, microarray analysis, Next Generation Sequencing analysis and any other Sequencing method such as, but not limited to, nanopore sequencing, third-generation sequencing, Sanger sequencing analysis, digital or droplet PCR, quantitative Real-time PCR analysis and/or PCR amplification of the nucleic acids derived from respective sample and/or a sequential or parallel combination of previously listed steps and techniques.
Herein, Next Generation Sequencing relates to, without being limited to, RNA sequencing, DNA sequencing, Whole Genome Sequencing, Whole Exome Sequencing, Single Cell Sequencing, Targeted Sequencing, NanoString Hyb & Seq NGS analysis, nCounter-analysis, sequencing methods allowing two-dimensional resolution of transcriptome or epigenome patterns in tissues such as, but not limited to, spatial sequencing or sequencing methods allowing the characterization of chromatin status such as, but not limited to, ATAC-Seq or ChIP-Seq.
Herein, the gene expression of said reference genes in said sample can be analyzed by, without being limited to, transcriptome-wide gene expression analysis comprising, without limitation to, the analysis techniques and methods of nucleic acids and nucleic acid sequences described herein. Herein, the gene expression of reference genes in said sample can be calculated, without being limited to, using the abundancies of target nucleic acids and/or target nucleic acid sequences in or obtained from a patient sample.
The score, herein also referred to as ProstaTrend score, may be calculated herein, without being limited to, by using the expression values of said target genes and/or the abundancies of each one or more target nucleic acid sequences in said sample or derived from said sample, which each possess an identical nucleic acid sequence to one of the genes, or of a transcript or isoform thereof, that are listed in Table 2.
In greater detail, in the herein disclosed method for predicting the outcome or classifying the disease of prostate carcinoma patients said score value for said patient may be calculated by, without being limited to, using the abundancies of said target nucleic acid sequences in said sample to calculate expression values for all corresponding reference genes in said sample, and wherein for each reference gene the abundancies of all target nucleic sequences that are identical to the nucleic acid sequence of at least one or more transcripts and/or isoforms of said reference gene are summarized for each reference gene, and wherein for each patient k the median over the weighted standardized gene expression values of all reference genes selected for said calculation gi is calculated applying:
scorek=Median(logHRi*gi), and
wherein the weight for each significant gene i is the estimated univariate log hazard ratio (logHR) from the Cox regression meta-analysis of a training cohort to account for the direction of the prognostic effect of said gene i as well as for the effect size, and
wherein a normalization of said gene expression values is done cohort-wise, if a set of samples from the same cohort is provided, or, in case of a single sample, an add-on normalization with respect to the normalized gene expression values of a training cohort is conducted, and wherein a negative value of said score characterizes said patient with lower risk of an adverse prognosis and/or death of disease and an increased chance of extended biochemical recurrence-free and/or death of disease-free survival, and a positive value of said score characterizes said patient with a higher risk of an adverse prognosis and/or death of disease and an decreased chance of extended biochemical recurrence-free and/or death of disease-free survival.
The herein disclose invention may also relate to a method for use in the treatment of prostate cancer, with the method comprising the steps of: a) obtaining a sample from a patient, wherein the sample comprises a plurality of nucleic acids, b) analyzing the plurality of nucleic acids, thereby determining the abundancy of at least one or more different target nucleic acid sequences, c) normalizing of the target nucleic acid sequence abundancies against a reference dataset and standardization, d) weighting the abundancy of the at least one or more different target nucleic acid using a corresponding reference value and calculating a score value, and wherein the outcome or classification of the disease of a prostate carcinoma patient from whom said sample was derived is predicted based on said score value, and wherein, if the score value is above a threshold value of 0, the patient is classified as having a high risk of death of disease and/or biochemical recurrence and/or if the score value is below a threshold value of 0, the patient is classified as having a low risk of death of disease and/or biochemical recurrence.
A flowchart depicting the number of patients included in the training cohort and reasons for exclusion according to REMARK (McShane et al., 2005; J Natl Cancer Inst.). For 164 patients who underwent RP fresh-frozen (FF) tumor specimens fulfilled all inclusion criteria and gene expression variation by custom gene expression microarrays (FF_array_RP) was assessed. A subset of 40 patients was selected to perform a strand-specific transcriptome-wide sequencing (FF_seq_RP). For 14 of those patients, differential gene expression was confirmed in an FFPE tissue specimen by strand-specific RNA-seq. All but one sample passed the quality filter and entered the meta-analysis (FFPE_seq_RP). Gene expression variation in FFPE biopsies was assessed for 25 patients from an independent cohort with strand-specific RNA-seq. Of those, 16 samples passed the quality filter and entered the meta-analysis (FFPE_seq_Bx). n denotes the number of patients for which tissue specimens were processed. The number of DoD (death because of disease) events (primary endpoint) included in the meta-analysis is presented for each cohort. RIN corresponds to the RNA integrity number.
Statistical analysis was carried out evaluating the performance of the continuous ProstaTrend score, relating to the herein described invention, distinguishing patients with and without BCR confirming a significant correlation of the invention (ProstaTrend score) with BCR (p=1.7e−05) and providing the estimation of the conditional probability of BCR at a given score. The figure depicts logistic regression analysis for the ProstaTrend score and BCR. Patients with BCR are plotted in grey scattered around the horizontal axis at 1. Patients without BCR are depicted in black at the bottom of the graph scattered around 0. The dashed line indicates the overall frequency of BCR in the “TCGA PRAD” patient cohort (13%). The logistic regression curve represents the estimated conditional probability of BCR at a given ProstaTrend score.
The present invention (ProstaTrend score) allows for significant discrimination of patients with low/intermediate and high Gleason score (GS; n=186), characterized by percentage of BCR-free survival.
The present invention (ProstaTrend score) allows for significant discrimination of different risk groups, characterized by percentage of “death of disease” (DoD)-free survival. The
(A) Representation of the “FF_array_RP” dataset using Self-organizing maps (R-package oposSOM v.1.16). The map illustrates the distribution of overexpressed sets of correlated genes (spots) across the SOM landscape in tumor samples. Spots that significantly enrich prognostic genes (i.e. genes listed in Table 2) after adjusting for multiple testing by Bonferroni correction (adjusted p-value<0.05) are color-coded and annotated. The adjusted p-values for the enrichment of prognostic genes are 1.1*10-104 for spot B, 1.9*10−18 for spot A, 3.4*10−10 for spot I, 8.7*10−08 for spot S, 1.2*10−06 for spot J and 2.1*10−03 for spot V. (B) Box plot of the logHRs for the prognostic genes of significantly enriched spots. (C) Spots enriched with prognostic genes are sorted according to the number of included prognostic genes. If genes of specific biological pathways were enriched, spots were assigned to a functional signature of gene ontology (GO) or gene-sets collected from publications and independent analyses that are part of the oposSOM package. The top five genes (ranked decreasingly by correlation with mean spot expression profile) are shown for each spot. “#” indicates genes only included in Prolaris, Oncotype and/or Decipher signatures but not in the ProstaTrend gene signature (Table 2). “##” indicates genes included in the ProstaTrend signature, but not in the Prolaris, Oncotype and/or Decipher signatures, and “###” indicates genes that are occurring in the ProstaTrend signature and in the Prolaris, Oncotype and/or Decipher signatures. (D) Bar plots, depicting the distribution of gene types (biotypes) among all genes in the spots, significantly enriched with prognostic genes (left panel) and the distribution of gene types (biotypes) included in the spot and the ProstaTrend score (right panel). The y-axis is square-root-scaled to ensure readability for small groups. (E) Kaplan-Meier curves for DoD for patients with mean expression of the genes in the respective spot>0 compared to patients with mean expression of genes in the respective spot 0. The prognostic value of the spots was assessed with a log-rank test. Ig/TcR: immunoglobulin/T-cell receptor; lncRNA: long non-coding RNA; small ncRNA: non-coding RNA.
The different scores were computed exactly as the herein disclosed prognostic ProstaTrend score but for different selections of ProstaTrend reference genes (selected from Table 2) and the calculated scores were applied to the “TCGA PRAD” patient cohort to predict BCR-free survival. Kaplan-Meier curves were truncated if the number of patients at risk dropped below ten in any arm. (A) The same Kaplan-Meier curve as presented in
AUC (area under curve) for classification of biochemical recurrence (yes/no) in the “TCGA_PRAD” validation cohort is shown. AUC is estimated for n=1 to n=1396 genes from the complete set of the herein disclosed gene set (Table 2), which is used to calculate the ProstaTrend score. For each set N=100 bootstrap samples (with replacement) were evaluated, AUC for classification of biochemical recurrence calculated and plotted. With increasing size of the gene set the AUC improves and the variation decreases. For roughly n˜[100-300] the AUC is ˜0.7 and the increase of AUC decreases.
AUC for classification of biochemical recurrence (yes/no) in this cohort is shown. AUC is estimated for n=1 to n=1396 genes from the complete set of the herein disclosed gene set (Table 2), which is used to calculate the ProstaTrend score. With increasing size of the gene set the AUC increases up to 0.68 at ˜n=400 genes (old FFPE specimen excluded).
For each reference gene (Table 2), a univariate fixed-effect meta-analysis of Cox-regression models in FF RP tissue (FF_array_RP and FF_seq_RP), FFPE RP tissue (FFPE_seq_RP) and FFPE biopsies (FFPE_seq_Bx) was conducted to identify genes with expression changes significantly associated with DoD. (A) An exemplary result of the conducted meta-analysis showing the output for the gene NUSAP1 (included in our prognostic score as well as in the Prolaris signature). (B) A histogram of p-values for the univariate fixed-effect meta-analysis of all genes included in the analysis. The histogram shows a strong enrichment for p-values near zero indicating an association with prognosis for a substantial fraction of genes. The estimated proportions of true and false null hypotheses by fitting a mixture model using the R-package fdrtool are 67.4% (null component: grey dotted line) and 32.6% (alternative component: solid black line), respectively. (C) Forest plot of the overall logHRs and corresponding 95% confidence intervals (95%-CI) for all genes included in the ProstaTrend score, i.e., all genes with tail-based false discovery rate (Fdr) <0.05 (n=1396). LogHRs are colored according to the particular gene biotype. For protein-coding genes, the y-axis is condensed to ensure readability for the smaller non-protein coding gene groups.
Table 1: Prognostic value of dichotomized ProstaTrend score compared to clinical features in an independent cohort (“TCGA PRAD”, n=332 samples) with the primary endpoint BCR.
The prognostic value of dichotomized ProstaTrend and clinical risk factors that are available for the “TCGA PRAD” patient cohort was assessed by Cox proportional hazard regression. We analyzed dichotomized ProstaTrend score values (>0 vs ≤0), i.e. assigned value 0 to a patient if the patients' ProstaTrend score≤0 and assigned value 1 to a patient if the patients' ProstaTrend score>0. We analysed PSA on the log-scale and age as a continuous variable. The GS was assessed based on RP tissue and Gleason grading groups (GG)≥3 was compared to GG <3. For resection status we compared R+vs R0. The upper part of the table shows univariate results, the lower part contains pairwise analysis of the dichotomized ProstaTrend score (see definition above) and each clinical feature. The bottom line shows the result of the cox regression including ProstaTrend, GG and pathological stage, i.e. all clinical features that remained significant in a pairwise model with the dichotomized ProstaTrend score.
Table 2: List of reference genes, herein also referred to as prognostic genes, used to calculate the ProstaTrend score according to the present invention.
To identify genes associated with DoD we applied weighted univariate Cox regression analysis on a set of 20,821 genes that passed the expression quality filter for all included training cohorts separately and used the estimated log hazard-ratios (logHR) and the respective standard errors (SEs) in a fixed-effect-model meta-analysis. For a conservative correction for multiple testing, we selected the 1,396 genes with a tail-based Fdr <0.05 as a marker set including 1,376 known genes and 20 novel genes.
The logHR values in the present table are used during the calculation of the ProstaTrend score as a weighting factor for the respective gene. After this step these weighted genes are combined by the median. Genomic position of genes is w.r.t. human genome assembly hg19.
Table 3: Samples of the validation patient cohort “TCGA PRAD”.
Table providing for each sample of the validation patient cohort (“TCGA PRAD”) the ID, the calculated ProstaTrend score, occurrence of biochemical recurrence within follow-up, and the follow-up time in days.
Table 4: Samples of the training patient cohort “FF_array_RP”.
Table providing for each sample of a training patient cohort (“FF_array_RP”) the ID, the calculated ProstaTrend score (with leave-one out cross validation), occurrence of DoD within follow-up, and follow-up time in years.
Table 5: Prognostic value of continuous ProstaTrend score compared to clinical features in an independent cohort (“TCGA PRAD”, n=332 samples) with the primary endpoint BCR.
We assessed the prognostic value of the continuous ProstaTrend score and clinical risk factors that are available for the “TCGA PRAD” cohort by Cox proportional hazard regression. We analyzed continuous ProstaTrend score, PSA on the log-scale and age as a continuous variable. The GS was assessed based on RP tissue and Gleason Grading Groups (GG) ≥3 was compared to GG <3. For resection status we compared R+vs R0. The upper part of the table shows univariate results, the lower part contains pairwise analysis of ProstaTrend and each clinical feature. The bottom line shows the result of the cox regression including ProstaTrend, GG and pathological stage, i.e. all clinical variables that remained significant in the pairwise analysis with ProstaTrend.
The present invention relates to a method (ProstaTrend) for predicting the outcome and/or classifying the disease of prostate carcinoma patients, wherein the method comprises calculating and applying a prognostic multi-gene expression score (ProstaTrend score) based on the expression of a set of genes that are significantly associated with prostate cancer prognosis (Table 2) in a patient sample. The herein disclosed method allows for molecular risk stratification and improved prediction of long-term patient prognosis in the treatment of prostate cancer. The invention further relates to a method, comprising the prediction of death of disease (DoD) and biochemical recurrence (BCR), as well as the prediction of a long-term prognosis and the disease outcome of prostate cancer patients. The method provided by the invention (ProstaTrend) is independent from clinical and histological parameters after diagnosis and facilitates the identification of high-risk patients among patients with low or intermediate histological-risk profiles prior and after surgery.
One of the surprising technical effects of the present invention is the gain of patient outcome prediction accuracy through the application of the ProstaTrend score method, which is shown herein by a concordance analysis (C-Index) for a clinical model based on grading groups, pathological stage and resection status with and without inclusion of the herein disclosed ProstaTrend score method for samples of the patient “TCGA PRAD” cohort (see also Example 3). With the inclusion of the ProstaTrend score method the concordance index increased by 4.9% [95%-CI: −0.3-8.6%] from 71.4% to 76.3%. In greater detail, the gain of prediction accuracy by using the herein disclosed ProstaTrend score, is shown by the comparison of the concordance index between two multivariable cox models, one featuring clinical parameters grading groups, pathological staging, and resection status only, while the second model also includes the herein disclosed ProstaTrend score method. 1000 bootstrap samples were generated using the R-Package boot v1.3-2418 while keeping the number of events constant to estimate the 95% confidence for the difference of the concordance index (C-index) between both models.
In one embodiment the present invention relates to a method for predicting the outcome or classifying the disease of a prostate carcinoma patient with the method comprising the steps of obtaining a biological sample, wherein the sample comprises a plurality of nucleic acids, a plurality of cells, a biopsy and/or human or animal tissue and obtaining a plurality of nucleic acids from the sample, followed by subsequent analysis of the plurality of nucleic acids, thereby determining the sequences of the nucleic acids from the sample and/or the abundancies of target nucleic acid sequences from the sample, followed by comparing the sequences of the nucleic acids and/or the abundancies of the target nucleic acid sequences derived from the sample to a set of known reference nucleic acid sequences and finally calculating a score (ProstaTrend score) value depending on the comparison of the target nucleic acid sequences and/or of the target nucleic acid sequence abundancies in relation to the set of known reference gene nucleic acid sequences, wherein a difference or similarity in at least one of the nucleic acid sequences and/or nucleic acid sequence abundancies between the known reference gene nucleic acid sequences and the target nucleic acid sequences from said sample is used for calculating a score value that is predictive of a prognosis and/or classification of the patient from whom the sample was derived.
In a further embodiment the present invention relates to a method for predicting the outcome or classifying the disease of a prostate carcinoma patient, the method comprising the steps of: a) obtaining a sample from said patient, wherein the sample comprises a plurality of nucleic acids, b) analyzing the plurality of nucleic acids, thereby determining the abundancy of at least one or more different target nucleic acid sequences, c) normalizing of the target nucleic acid sequence abundancies against a reference dataset and standardization, d) weighting the abundancy of the at least one or more different target nucleic acid using a corresponding reference value and e) calculating a score value, and wherein, if the score value is above a threshold value of 0, the patient is classified as having a high risk of death of disease and/or biochemical recurrence and/or if the score value is below a threshold value of 0, the patient is classified as having a low risk of death of disease and/or biochemical recurrence.
In another embodiment the invention relates to a method for predicting the outcome or classifying the disease of a prostate carcinoma patient with the method comprising the steps of: a) obtaining a biological sample from said patient, wherein the sample comprises a plurality of nucleic acids, a plurality of cells, a biopsy and/or human or animal tissue and obtaining a plurality of nucleic acids from the sample, b) analyzing the plurality of nucleic acids, thereby determining the sequences of the nucleic acids from the sample and/or the abundancy of at least one or more different target nucleic acid sequences from the sample, c) comparing the sequences of the nucleic acids and/or the abundancy of at least one or more different target nucleic acid sequences derived from the sample to a reference abundancy, thereby calculating a score value, wherein, if the score value is above a threshold value of 0, the patient is classified as having a high risk of death of disease and/or biochemical recurrence and/or if the score value is below a threshold value of 0, the patient is classified as having a low risk of death of disease and/or biochemical recurrence.
Herein, the patient from whom said sample was derived may be classified as having a high risk of death of disease and/or biochemical recurrence, if the calculated score value is above a threshold value of 0 (score value >0) and if the score value is below a threshold value of 0 (score value<0), the sample is classified as having a low risk of death of disease and/or biochemical recurrence.
In one embodiment of the invention, each target nucleic acid sequence derived from the analysis of the plurality of nucleic acids from said patient sample is identical to the nucleic acid sequence of a corresponding specific transcript and/or isoform of a reference gene selected from the group of genes associated with cancer progression or cancer malignancy. In another embodiment, said group of reference genes are associated with the progression or malignancy of prostate cancer. In a preferred embodiment, each target nucleic acid sequence derived from the analysis of the plurality of nucleic acids from said patient sample is identical to the nucleic acid sequence of a corresponding specific transcript and/or isoform of a reference gene selected from the group of genes listed in Table 2.
In other words, herein, said target nucleic acids or target nucleic acid sequences may be those nucleic acids or nucleic acid sequences derived from or present in said sample, that have an identical nucleic acid sequence to at least one or more reference genes, or at least one or more isoforms or transcripts thereof, selected from the group of genes listed in Table 2.
In one embodiment, said reference abundancy refers to the reference abundancy of the reference gene or reference gene nucleic acid sequence, wherein said abundancy of the target nucleic acids is weighted_by the logHR value of the corresponding reference gene listed in Table 2, wherein the nucleic acid sequence of said reference gene, or of a transcript or isoform thereof, is identical to the nucleic acid sequence of one or more of said target nucleic acid sequences derived from or present in said sample.
In one embodiment of the present invention a reference dataset may be, without being limited to, a dataset comprising the gene expression values and/or logHR values and/or further characterizing values of all or a group of genes comprising all or a fraction of said reference genes listed in Table 2, the dataset disclosed in Table 2 or a fraction of said dataset, a dataset comprising the gene expression values of genes associated with cancer or prostate cancer or another dataset comprising gene expression values.
In one embodiment of the invention, said corresponding reference value for weighting abundancies is a logHR value selected from the group of logHR values listed in Table 2 corresponding to one or more reference genes listed in Table 2, wherein the nucleic acid sequence of said one or more reference gene, or of a transcript or isoform thereof, is identical to the nucleic acid sequence of one or more of said target nucleic acid sequences from said sample.
In another embodiment, said logHR values refer to one or more logHR values from the group of logHR values listed in Table 2, wherein said one or more logHR values in Table 2 are used during the calculation of the ProstaTrend score as a weighting factor for the respective reference gene listed in Table 2 and/or the expression value or abundancy of said respective reference gene in said sample, wherein subsequently during the calculation these weighted reference genes or weighted reference gene abundancies or expression values are combined by the median, and wherein the nucleic acid sequence of said reference gene, or of a transcript or isoform thereof, is identical to the nucleic acid sequence of one or more of said target nucleic acid sequences derived from or present in said sample.
Herein, said genes disclosed in Table 2, are also referred to as reference genes, marker genes or prognostic genes.
In one embodiment of the invention, each logHR value from said logHR values listed in Table 2 corresponds to one or more reference genes from the group of reference genes listed in Table 2, wherein the nucleic acid sequence of said reference gene, or of a transcripts or isoforms thereof, is identical to the nucleic acid sequence of one or more of said target nucleic acid sequences from said sample, if said reference gene is expressed in said sample. In one embodiment of the invention said logHRs are weighting factors from the training of the prognosticator, that are fixed for every new sample, wherein the comparison to a reference comes due to add-on normalization (or combined normalization) with said established ProstaTrend reference gene set.
In another embodiment said target abundancy is weighted_by the logHR value of a corresponding reference gene listed in Table 2, wherein the nucleic acid sequence of said reference gene, or of a transcripts or isoforms thereof, is identical to the nucleic acid sequence of one or more of said target nucleic acid sequences from said sample.
Herein, each gene from the group of reference genes listed in Table 2, or one or more isoforms or transcripts thereof, that have a nucleic acid sequence identical to one or more target nucleic acid sequences from said sample, may herein be referred to, without being limited to, as corresponding reference gene to said target nucleic acid sequence. In other words, if a reference gene, listed in Table 2, is expressed in said patient sample, a target nucleic acid corresponding to said reference gene may be detected during the nucleic acid analysis of said sample and it's abundancy, herein also referred to as gene expression, may be determined. Therefore, the expression value and/or abundancy of each gene, listed in Table 2 and/or each transcript or isoform thereof, in said patient sample, may be determined, without being limited to, by analyzing the nucleic acid sequences and their abundancies in said sample. Said gene expression values of the selected reference genes in said sample may subsequently be used to, without being limited to, calculate a ProstaTrend score for the patient said sample was derived from, according to the method described herein.
Herein, a biological sample may be derived from, without being limited to, a human patient, a cell culture, an “organ-on-a-chip” model, a microphysiological system or an animal. Herein, a biological sample may be an organ, a tissue, a biopsy, a liquid biopsy, a blood sample, a sample from a Patient Derived Xenograft, a sample from an “organ-on-a-chip” model, a sample from a microphysiological system, a cell line or cell culture of a patient-derived sample of a human patient, an animal or an animal model, of a cell line or of an organoid. In a more preferred embodiment of the invention the sample is an organ, a tissue, a biopsy, a liquid biopsy, a blood sample, a sample from an “organ-on-a-chip” model or a patient-derived cell culture of a human patient. In the most preferred embodiment, the sample is a tissue, a biopsy or a liquid biopsy sample of a human patient. In one embodiment the biopsy sample is a fresh sample, a frozen sample or a FFPE preserved sample.
Examples of methods for nucleic acid sequence analysis are known to those of ordinary skill in the art and may herein comprise, without being limited to, optional lysis of the tissue and/or the cells and optional subsequent purification of respective nucleic acids followed by optional fragmentation, enzymatic digestion or ligation of respective nucleic acids to anchors, solid phases or barcodes, reverse transcription, microarray analysis, Next Generation Sequencing analysis and any other Sequencing method such as, but not limited to, nanopore sequencing, third-generation sequencing, Sanger sequencing analysis, NanoString Hyb & Seq NGS analysis, nCounter-analysis, digital or droplet PCR analysis, quantitative Real-time PCR analysis and/or PCR amplification of the nucleic acids derived from respective sample and/or a sequential or parallel combination of two or more of the previously listed steps and techniques.
Herein, Next Generation Sequencing relates to, without being limited to, RNA sequencing, DNA sequencing, Whole Genome Sequencing, Whole Exome Sequencing, Single Cell Sequencing, Targeted Sequencing, NanoString Hyb & Seq NGS analysis, nCounter-analysis, sequencing methods allowing two-dimensional resolution of transcriptome or epigenome patterns in tissues such as, but not limited to, spatial sequencing or sequencing methods allowing the characterization of chromatin status such as, but not limited to, ChIP-Seq or ATAC-Seq.
In one embodiment of the invention the term patient can be equivalent, but is not limited, to a human patient, a Patient Derived Xenograft, an “organ-on-a-chip” model, a microphysiological system, an animal, an animal model, a cell line or a cell culture of a patient-derived sample of a human patient, of an animal or of an animal model, of an “organ-on-a-chip” model, of a microphysiological system, of a cell line or of an organoid. In a preferred embodiment the patient is a human patient, a Patient Derived Xenograft or an animal. In an even more preferred embodiment of the invention the patient is a human patient.
As used herein, ProstaTrend refers to a method that can be used in, without being limited to, the treatment of prostate cancer for molecular risk stratification and improved prediction of long-term prognosis for prostate cancer patients with the method comprising calculating and applying a prognostic multi-gene expression score, which is also referred to herein as ProstaTrend score, based on the expression values of a set of genes that are significantly associated with prostate cancer prognosis (Table 2), herein also referred to as reference genes. The ProstaTrend method disclosed herein can also be used in the prediction of death of disease (DoD) and biochemical recurrence (BCR), or a probability of one or both, or of a long-term prognosis and/or in the prediction of outcome for prostate cancer patients.
In one embodiment of the invention the score (ProstaTrend score) obtained by the method described herein can be used in the treatment of prostate cancer for, without being limited to, calculating a prognosis and/or a classification and/or stratification of prostate carcinoma patients from whom respective biological sample was derived. In another embodiment of the invention the score obtained by the method described herein can be used for the classification and/or grading of respective analyzed sample.
In another embodiment of the invention the score obtained by the method described herein can be used in the treatment of prostate cancer to, without being limited to, classify the aggressiveness of a prostate carcinoma during initial diagnosis, before, during or after treatment of respective cancer. In a further embodiment of the invention the score obtained by the method described herein can be used to calculate the stage of and/or to classify the aggressiveness of metastases of a prostate carcinoma during initial diagnosis, before, during or after cancer treatment. The present invention further relates to a method for classifying patients according to their gene expression profile into a numeric score. Herein, this score classifies patients associated with adverse prognosis and reduced DoD-free survival or BCR-free survival with a score value of >0. A score value of 0 further classifies patients with an extended DoD-free or biochemical recurrence-free survival (
In one embodiment the score obtained by the herein described method can be used in the treatment of cancer, such as, but not limited to, prostate cancer, prostate carcinomas or metastases of prostate carcinomas, to support and improve the decision whether adjuvant therapy after surgery, such as RP, should be scheduled, as well as to support and improve the choice of therapy for patients after initial diagnosis, during active surveillance and after focal, organ-preserving therapies. Herein “therapy” includes, but is not limited to, surgery, radical prostatectomy (RP), “active surveillance” or “watchful waiting”, radiotherapy, focal therapies, and wherein “focal therapy” includes, but is not limited to, vascular targeted photodynamic therapy, such as, but without being limited to, TOOKAD, High Intensity Focused Ultrasound (HiFu) and irreversible Electroporation (IRE), such as, but without being limited to, NanoKnife and Focal Cryoablation.
In one embodiment of the present invention, when correlating the score, obtained by the method described herein, to the proportion of prostate carcinoma patients developing BCR, said score of a value of −0.2 relates to an estimated conditional probability of BCR of respective prostate carcinoma patient of about 5%, whereas said score of a value of 0 relates to an estimated conditional probability of BCR of about 11% and said score of a value of 0.2 relates to an estimated conditional probability of BCR of about 23% (Table 3;
In another embodiment the herein disclosed method and ProstaTrend score relate to a method for predicting the outcome or classifying and/or stratifying the disease of prostate carcinoma patients and is calculated by combining the expression values of all selected genes in a patient-wise prognostic score.
In one embodiment of the invention the method comprises the detection of nucleic acids that have a target nucleic acid sequence identical to one or more of said reference genes, or one or more transcripts or isoforms thereof, listed in Table 2, followed by the determination or measurement of the abundancy of each target nucleic acid sequence in said sample, wherein said abundancy is used to calculate the gene expression value of said corresponding reference gene from the group of reference genes listed in Table 2 in said sample.
In one embodiment of the invention, said score can be calculated by the herein disclosed method, wherein the abundancies in said sample of all known isoforms or transcripts of each of said reference genes listed in Table 2 are summarized for each reference gene during the calculation.
In another embodiment the invention relates to a method for predicting the outcome or classifying the disease of prostate carcinoma patients, the method comprising the steps of: a) obtaining a sample from said patient, wherein the sample comprises a plurality of nucleic acids, b) analyzing the plurality of nucleic acids, thereby determining the abundancy of at least two one or more different target nucleic acid sequences, c) normalizing said target nucleic acid sequences against a reference dataset and standardization, d) weighting the abundancy of the at least one or more different target nucleic acid using the respective log hazard ratios presented in Table 2, and e) calculating a score value by summarizing the abundancies of all known isoforms or transcripts of each of said genes listed in Table 2, wherein, if the score value is above a threshold value of 0, the patient is classified as having a high risk of death of disease and/or biochemical recurrence and/or if the score value is below a threshold value of 0, the patient is classified as having a low risk of death of disease and/or biochemical recurrence.
In detail, in one embodiment of the invention the disclosed method of calculating a predictive score value (ProstaTrend score) for use in predicting the outcome or in classifying the disease of a prostate carcinoma patient comprises the calculation of expression values for at least one reference gene that were detected to be expressed in the sample, based on the abundancies of corresponding target nucleic acid sequences in said sample, wherein the abundancies of all target nucleic acid sequences, which are identical to the sequence of one or more isoforms and/or transcripts of a reference gene are summarized for each reference gene respectively, and wherein for each patient k the median over the weighted standardized gene expression values of all genes selected for said calculation g; is calculated applying:
ProstaTrendk=Median(logHRi*gi),
wherein the weight for each significant gene i is the estimated univariate log hazard ratio (logHR) from the Cox regression meta-analysis to account for the direction of the prognostic effect of gene i as well as for the effect size, wherein a negative ProstaTrend-score (score value<0) characterizes patients with lower risk of an adverse prognosis and an increased chance of extended BCR-free and/or DoD-free survival, while a positive score (score value >0) characterizes patients with an increased risk of an adverse prognosis and a lower chance of extended BCR-free and/or DoD-free survival.
Briefly, in one embodiment of the invention the score (ProstaTrend score) is calculated herein by using the expression abundancies of each one or more target nucleic acid sequences from a sample that have an identical sequence to the sequence of one of the genes listed in Table 2 or one of the isoforms or transcripts of said genes, and wherein the score is calculated by summarizing the abundancies of all known isoforms or transcripts of each of said genes listed in Table 2.
In one embodiment of the invention said score value is calculated by summarizing the abundancies of all known isoforms or transcripts of each of said reference genes listed in Table 2.
In one embodiment of the invention, the ProstaTrend score for a patient is calculated as the median of the weighted, normalized and standardized gene expression values according to the before mentioned method using a set of at least 50 or more prognostic genes, herein also referred to as reference genes, which are disclosed herein in Table 2.
In one embodiment of the present invention said set of reference genes used to calculate the ProstaTrend score comprises at least 10, 50, 100, 150, 200, 250, 300, 350, 400, 450, 500, 750, 1000, 1250 or 1396 of the genes disclosed in Table 2. In another embodiment said set of reference genes comprises a number within the range of 50-1396 of the genes disclosed in Table 2. In another embodiment of this invention said set of reference genes used to calculate the ProstaTrend score comprises a number within the range of 50-500 of the genes disclosed in Table 2. In a preferred embodiment of this invention said set of reference genes used to calculate the ProstaTrend score comprises a number within the range of 100-500 (
In one embodiment of the invention the normalization of gene expression values is done cohort-wise, if a set of samples from the same patient cohort is provided, or, in case of a single sample, an add-on normalization with respect to the normalized gene expression values of the training cohort is conducted. Further, herein the expression value of each gene, herein also referred to as reference gene, is weighted by the log hazard ratio (logHR) from the Cox regression meta-analysis of the training cohort. Herein, the weights for each gene are provided in Table 2.
In one embodiment, the present invention relates to a method for use in the treatment of prostate cancer, the method comprising the steps of: a) obtaining a sample from a patient, wherein the sample comprises a plurality of nucleic acids, b) analyzing the plurality of nucleic acids, thereby determining the abundancy of at least one or more different target nucleic acid sequences, c) normalizing of target nucleic acid sequences or target nucleic acid sequence abundancies against a reference dataset and standardization, d) weighting the abundancy of the at least one or more different target nucleic acid using the respective log hazard ratios presented in Table 2 and e) calculating a score value, and wherein, if the score value is above a threshold value of 0, the patient is classified as having a high risk of death of disease and/or biochemical recurrence and/or if the score value is below a threshold value of 0, the patient is classified as having a low risk of death of disease and/or biochemical recurrence.
In another embodiment the herein disclosed method relates to a method for use in the treatment of prostate cancer, the method comprising the steps of: a) obtaining a sample from a patient, wherein the sample comprises a plurality of nucleic acids, b) analyzing the plurality of nucleic acids, thereby determining the abundancy of at least one or more different target nucleic acid sequences, c) normalizing of the target nucleic acid sequence abundancies against a reference dataset and standardization, d) weighting the abundancy of the at least one or more different target nucleic acid using a corresponding reference value and e) calculating a score value, and wherein the outcome or classification of the disease of a prostate carcinoma patient from whom said sample was derived is predicted based on said score value, and wherein, if the score value is above a threshold value of 0, the patient is classified as having a high risk of death of disease and/or biochemical recurrence and/or if the score value is below a threshold value of 0, the patient is classified as having a low risk of death of disease and/or biochemical recurrence.
In another embodiment of said method for use in the treatment of prostate cancer, the outcome or classification of the disease of a prostate carcinoma patient from whom said sample was derived is predicted based on said score value.
The embodiments discussed herein with reference to the method for predicting the outcome or classifying the disease of a prostate carcinoma patient apply also to the method for use in the treatment of prostate cancer.
To identify genes associated with DoD four clinical cohorts (
The herein disclosed list of reference genes (Table 2), which can be used to calculate the herein disclosed prognostic ProstaTrend score, is consistent with already established prognostic gene signatures. In total, 24 genes of the Decipher, Oncotype, and Prolaris signatures were included in the list of genes disclosed by the present invention, which can be used to calculate the herein disclosed prognostic ProstaTrend score. Removing these genes did not affect the accuracy of the calculated ProstaTrend score (
The ProstaTrend score, described by the current invention, admitted high discrimination accuracy in the “FF_array_RP” patient cohort (
Patients with score values >0 had significantly reduced prostate cancer-specific survival as compared to patients with score values ≤0 (5-year survival: 83% vs. 98%; log-rank test: p=0.0002). The ProstataTrend score remained an independent predictor for the time to DoD in a multivariate Cox regression after adjusting for Gleason grading group (GG) 3 and positive resection status (
To confirm the prognostic relevance of the continuous ProstaTrend score, we assessed the concordance with the order of DoD events. A univariable analysis showed a C-index of 78.5% [95%-CI: 68.5-85.9%] for the cross validated ProstaTrend score.
In order to supply a strong validation study, the effectiveness of the ProstaTrend score, disclosed by the present application, in an external patient cohort (TCGA PRAD) was demonstrated. The ProstaTrend score allowed for significant discrimination of different risk groups (
In univariate Cox regression analysis, the dichotomized ProstaTrend score (dichotomized, for example dividing the ProstaTrend score in two separate classes: value 1 in case of a ProstaTrend score value >0 and value 0 in case of a ProstaTrend score value ≤0) was the most significant predictor for BCR compared to log(PSA), GG ≥3, pathological stage ≥T3, lymph node status, age and resection status. A pairwise analysis with the ProstaTrend score and each clinical feature resulted in the statistical significance of the ProstaTrend score in all models. In each paired model, the ProstaTrend score was more significant than the clinical feature and the ProstaTrend score remained significant when adjusted for GG and stage, for example all clinical variables that remained significant in the pairwise analysis (see Table 1).
Similarly, in univariate Cox regression analysis the continuous ProstaTrend score was the most significant predictor for BCR compared to log(PSA), GG ≥3, pathological stage≥T3, age and resections status. A pairwise analysis with ProstaTrend and each clinical feature resulted in the statistical significance of ProstaTrend in all models. ProstaTrend remained significant when adjusted for GG and stage, i.e. all clinical variables that remained significant in the pairwise analysis (see Table 5).
To assess the prognostic potential of the gene expression signature in a low and intermediate risk group the ProstaTrend score was applied to patients with Gleason score (GS) ≤7 (
To investigate the gain of prediction accuracy through the application of the ProstaTrend score, we applied the concordance analysis for a clinical model based on GG, pathological stage and resection status with and without inclusion of the ProstaTrend score. With the inclusion of the ProstaTrend score the concordance index increased by 4.9% [95%-CI: −0.3-8.6%] from 71.4% to 76.3%.
To investigate the functional context of genes associated with prognosis in prostate cancer unsupervised clustering with self-organizing maps (SOM) was performed. OposSOM v.1.1619 was applied to the quality-filtered and between-array normalized genes of the FF_array_RP patient cohort. 23 correlated overexpression spots were identified (
The list of genes disclosed by the present application (Table 2) that is used for the calculation of a ProstaTrend score comprises genes of all gene sets representing the biological processes of the “hallmarks of cancer” (Hanahan et al., 2011; Cell and Loeffler-Wirth et al, 2019; Genome Med.). In contrast, Decipher, Oncotype, and Prolaris had no overlap with at least three hallmark gene sets indicating that the spectrum of biological characteristics covered by these assays is limited compared to the ProstaTrend score. The “hallmarks of cancer” have been defined as a set of functional capabilities that a tumor acquires during tumor development (Hanahan et al., 2011; Cell). We compared previously defined gene sets referring to these “hallmarks” (Loeffler-Wirth et al, 2019; Genome Med.) to determine the overlap with the herein disclosed gene set (Table 2) used to calculate the ProstaTrend score and thus the potential of the ProstaTrend score to represent these biological processes. All “hallmarks” were represented within the herein disclosed gene set (Table 2), which is used herein to calculate the ProstaTrend score, for multiple of genes of said gene set: angiogenesis (n=17), controlling genomic instability (n=50), glucose energetics (n=3), inflammation (n=18), invasion and metastasis (n=71), proliferation (n=156), replicative immortality (n=57) and resisting death (n=45) suggesting that the herein disclosed ProstaTrend score covers the full spectrum of biological characteristics involved in tumorigenesis.
| Number | Date | Country | Kind |
|---|---|---|---|
| 20170321.2 | Apr 2020 | EP | regional |
This application is a national stage application under 35 U.S.C. § 371 of International Application No. PCT/EP2021/060092, filed on Apr. 19, 2021, which claims the benefit of European Patent Application No. 20170321.2, filed on Apr. 20, 2020, The content of these earlier filed applications are hereby incorporated herein by reference in their entirety.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/EP2021/060092 | 4/19/2021 | WO |
