Machine Learning-Assisted Decision Making in Orthopaedic Oncology.

¼ Artificial intelligence is an umbrella term for computational calculations that are designed to mimic human intelligence and problem-solving capabilities, although in the future, this may become an incomplete definition. Machine learning (ML) encompasses the development of algorithms or predictive models that generate outputs without explicit instructions, assisting in clinical predictions based on large data sets. Deep learning is a subset of ML that utilizes layers of networks that use various inter-relational connections to define and generalize data.¼ ML algorithms can enhance radiomics techniques for improved image evaluation and diagnosis. While ML shows promise with the advent of radiomics, there are still obstacles to overcome.¼ Several calculators leveraging ML algorithms have been developed to predict survival in primary sarcomas and metastatic bone disease utilizing patient-specific data. While these models often report exceptionally accurate performance, it is crucial to evaluate their robustness using standardized guidelines.¼ While increased computing power suggests continuous improvement of ML algorithms, these advancements must be balanced against challenges such as diversifying data, addressing ethical concerns, and enhancing model interpretability.

Peak Creatinine, Cardiopulmonary Bypass, and Mortality After Stage 1 Single-Ventricle Reconstruction.

BACKGROUND: Serum creatinine is the most commonly used marker to diagnose acute kidney injury. Studies exploring creatinine patterns in the single-ventricle population are scarce. We studied serum creatinine up to 5 postoperative days after the stage 1 operation and assessed its relationship with outcomes. METHODS: Neonates who underwent a first-stage single-ventricle operation (Norwood or a Damus-Kaye-Stansel) between 2005 and 2017 were retrospectively analyzed. Peak percentage creatinine change (PPCC) was defined as the difference between the baseline (preoperative) and the peak postoperative level (within 5 postoperative days), expressed as a percentage of the baseline level. RESULTS: Among 187 neonates included, the median PPCC was 38.7% (interquartile range, 14.1%-73.1%), and in-hospital mortality was 17% (31 of 187). A controlled analysis showed that for every 10-minute increase in cardiopulmonary bypass duration (CPB), the PPCC increased by 1.8% (95% confidence interval [CI], 0.7%-2.9%; P = .002). Risk of in-hospital death increased log-linearly with PPCC. The adjusted odds ratios for death in the hospital associated with a 50%, 100%, and 200%, increase in peak percentage creatinine change were 1.85 (95% CI, 1.23-2.78), 3.41 (95% CI, 1.15-7.72), and 11.66 (95% CI, 2.28-59.63), respectively. In-hospital death was also associated with CPB duration (adjusted odds ratio, 1.13 per 10-minute increase; 95% CI, 1.05-1.22; P = .001). CONCLUSIONS: Increase in CPB duration has a strong linear association with increase in PPCC after stage 1 single-ventricle reconstruction. Increase in PPCC and CPB duration has a strong linear association with hospital mortality. It is important to identify therapies that minimize complications associated with prolonged CPB duration in high-risk populations.

Alpha-enolase is a potential prognostic marker in clear cell renal cell carcinoma.

Clear cell renal cell carcinoma (ccRCC) is an aggressive disease with unpredictable behaviour. Clinical parameters are not always accurate for prognosis prediction. The integration of molecular markers to prognostic models can significantly improve prognostic assessment and consequently patient management. We assessed the expression of alpha-enolase (ENO1) protein by immunohistochemistry in 360 patients with primary ccRCC and correlated its expression with multiple clinicopathological parameters including stage, grade, tumor size, disease-free and overall survival. Cox proportional hazard regression models adjusted for clinicopathological factors were used to test for a link between ENO1 expression and both disease-free and overall survival. We correlated ENO1 mRNA expression with overall survival in an independent set of 428 ccRCC cases from The Cancer Genome Atlas. ENO1 showed cytoplasmic, membranous and nuclear staining patterns. There is a statistically significant negative correlation between ENO1 expression, tumor stage, and grade. ENO1 expression also shows a statistically significant direct correlation with disease-free survival (p = 0.011) and overall survival (p = 0.030) in ccRCC. Patients with higher ENO1 expression had lower hazard ratio of recurrence, although this was not statistically significant (HR = 0.330, p = 0.060). These findings were validated at the mRNA level in an independent set of 428 ccRCC cases which also showed that low ENO1 expression is associated with significantly shorter overall survival. Down-regulation of ENO1 can be a predictor of poor prognosis in ccRCC, and it can be a potential prognostic marker.

miR-210 is a prognostic marker in clear cell renal cell carcinoma.

Accurate assessment of prognosis of clear cell renal cell carcinoma (ccRCC) is key in optimizing management plans to fit individual patient needs. miRNAs are short noncoding single-stranded RNAs that control the expression of target genes and may act as cancer biomarkers. We analyzed the expression of miR-210 in 276 cases of primary ccRCC and compared its expression in 40 pairs of adjacent normal and cancerous tissues. We assessed its expression in primary and metastatic tumors, in the common RCC subtypes, and the benign oncocytoma. The results were validated with an independent data set from The Cancer Genome Atlas. miR-210 was significantly overexpressed in ccRCC compared with normal kidney. miR-210(+) patients had a statistically higher chance of disease recurrence [hazard ratio (HR), 1.82; P = 0.018] and shorter overall survival (HR, 2.46; P = 0.014). In multivariate analysis, miR-210 lost its statistically significant association with shorter disease-free survival and overall survival after adjusting for tumor size and tumor, node, metastasis stage. Papillary RCC showed comparable miR-210 overexpression, whereas decreased up-regulation was seen in chromophobe RCC and oncocytoma. A number of predicted targets that might be involved in carcinogenesis and aggressive tumor behavior were identified. miR-210 is a potential therapeutic target and independent marker of poor prognosis of ccRCC.

Lactate dehydrogenase A is a potential prognostic marker in clear cell renal cell carcinoma.

BACKGROUND: Over 90% of cancer-related deaths in clear cell renal cell carcinoma (RCC) are caused by tumor relapse and metastasis. Thus, there is an urgent need for new molecular markers that can potentiate the efficacy of the current clinical-based models of prognosis assessment. The objective of this study is to evaluate the potential significance of lactate dehydrogenase A (LDHA), assessed by immunohistochemical staining, as a prognostic marker in clear cell renal cell carcinoma in relation to clinicopathological features and clinical outcome. METHODS: We assessed the expression of LDHA at the protein level, by immunohistochemistry, and correlated its expression with multiple clinicopathological features including tumor size, clinical stage, histological grade, disease-free and overall survival in 385 patients with primary clear cell renal cell carcinoma. We also correlated the LDHA expression with overall survival, at mRNA level, in an independent data set of 170 clear cell renal cell carcinoma cases from The Cancer Genome Atlas databases. Cox proportional hazards models adjusted for the potential clinicopathological factors were used to test for associations between the LDHA expression and both disease-free survival and overall survival. RESULTS: There is statistically significant positive correlation between LDHA level of expression and tumor size, clinical stage and histological grade. Moreover, LDHA expression shows significantly inverse correlation with both disease-free survival and overall survival in patients with clear cell renal cell carcinoma. Our results are validated by examining LDHA expression, at the mRNA level, in the independent data set of clear cell renal cell carcinoma cases from The Cancer Genome Atlas databases which also shows that higher lactate dehydrogenase A expression is associated with significantly shorter overall survival. CONCLUSION: Our results indicate that LDHA up-regulation can be a predictor of poor prognosis in clear cell renal cell carcinoma. Thus, it represents a potential prognostic biomarker that can boost the accuracy of other prognostic models in patients with clear cell renal cell carcinoma.

The chromatin remodeling gene ARID1A is a new prognostic marker in clear cell renal cell carcinoma.

Clear cell renal cell carcinoma (ccRCC) is the most common tumor of the adult kidney, with an increasing rate of incidence. Recently, exome sequencing studies have revealed that the SWI/SNF (switch/sucrose nonfermentable) members PBRM1 and ARID1A are mutated in ccRCC, and it has also been suggested that aberrant chromatin regulation is a key step in kidney cancer pathogenesis. Herein, we show that down-regulation of another SW/SNF component, ARID1A, occurs frequently in ccRCC. We detected copy number loss of ARID1A in 16% of patients with ccRCC. Immunohistochemistry indicated that 67% of ccRCC (53 of 79) had significantly lower expression of BAF250a, the protein product of ARID1A, than did the matched normal kidney cortex. In parallel, we conducted in silico mRNA expression analysis on 404 ccRCC tumors and 167 normal kidney cortex samples using publicly available databases and confirmed significant down-regulation of ARID1A in 68.8% of patients. We also show that decreased BAF250a protein and ARID1A mRNA expression correlate with tumor stage and grade. Our results indicate that both the protein and mRNA levels of ARID1A are statistically significant prognostic markers for ccRCC. Even after controlling for other confounders in the multivariate analysis, BAF250 retained its prognostic significance. BAF250a IHC is easy to perform and represents a potential biomarker that could be incorporated in laboratory practice to enhance the accuracy of the existing prognostic models.

Multilevel whole-genome analysis reveals candidate biomarkers in clear cell renal cell carcinoma.

Renal cell carcinoma (RCC) is the most common neoplasm of the kidney. We conducted an integrated analysis of copy number, gene expression (mRNA and miRNA), protein expression, and methylation changes in clear cell renal cell carcinoma (ccRCC). We used a stepwise approach to identify the most significant copy number aberrations (CNA) and identified regions of peak and broad copy number gain and loss, including peak gains (3q21, 5q32, 5q34-q35, 7p11, 7q21, 8q24, 11q13, and 12q14) and deletions (1p36, 2q34-q37, 3p25, 4q33-q35, 6q23-q27, and 9p21). These regions harbor novel tumor-related genes and miRNAs not previously reported in renal carcinoma. Integration of genome-wide expression data and gene set enrichment analysis revealed 75 gene sets significantly altered in tumors with CNAs compared with tumors without aberration. We also identified genes located in peak CNAs with concordant methylation changes (hypomethylated in copy number gains such as STC2 and CCND1 and hypermethylated in deletions such as CLCNKB, VHL, and CDKN2A/2B). For other genes, such as CA9, expression represents the net outcome of opposing forces (deletion and hypomethylation) that also significantly influences patient survival. We also validated the prognostic value of miRNA let-7i in RCCs. miR-138, located in chromosome 3p deletion, was also found to have suppressive effects on tumor proliferation and migration abilities. Our findings provide a significant advance in the delineation of the ccRCC genome by better defining the impact of CNAs in conjunction with methylation changes on the expression of cancer-related genes, miRNAs, and proteins and their influence on patient survival.

Integrated genomic characterization of the kallikrein gene locus in cancer.

BACKGROUND: The kallikrein-related peptidases (KLKs) have been implicated in many types of cancer, including prostate and ovarian. MATERIALS AND METHODS: We performed a comprehensive in silico study to characterize the KLK locus using transcriptomic (gene expression) and genomic (mutations and DNA copy number) data in prostate cancer (n=194), serous ovarian cancer (n=506), glioblastoma multiforme (n=206), and sarcoma (n=207) from The Cancer Genome Atlas and independent publicly available datasets to assess KLKs as cancer biomarkers. RESULTS: Overall, there was mRNA overexpression in prostate and serous ovarian cancer and decreased expression in glioblastoma multiforme. There was higher frequency of genomic loss in serous ovarian cancer, and rare KLK gene mutations observed in serous ovarian cancer and GBM. Dysregulation of KLKs correlates with survival: for prostate cancer, a combination of dysregulation of KLK1, 5 and 13 was associated with worse disease-free survival. CONCLUSION: We conclude that specific dysregulation of KLKs at the genetic and transcriptomic levels have useful prognostic value.

Pleiotropic action of renal cell carcinoma-dysregulated miRNAs on hypoxia-related signaling pathways.

The von Hippel-Lindau (VHL) gene is lost in ≈ 70% of all renal cell carcinomas (RCCs); however, increasing evidence supports the involvement of alternative mechanisms in the regulation of VHL expression, including suppression by microRNAs (miRNAs). miRNAs are small, noncoding RNA molecules that regulate gene expression through binding to target mRNAs. In this study, we found that miRNAs, which are dysregulated in cases of RCC, can target multiple members of RCC-related signaling pathways. Importantly, both VHL and the hypoxia-inducible factor 1-α gene are experimentally validated and are likely direct targets of miR-17-5p and miR-224, as shown by both luciferase assay and Western blot analysis. We found a negative correlation between miR-17-5p and its two predicted targets, VEGF-A and EGLN3, and between miR-224 and its targets SMAD4 and SMAD5 in RCC specimens, suggesting that downstream signaling pathways are also modulated by clear cell RCC-dysregulated miRs. Results from our bioinformatics analysis show that a single miRNA molecule can target multiple components of the same pathway and that multiple miRNAs can target the same molecule. Our results also indicate that miRNAs represent a mechanism for the inactivation of VHL in cases of RCC and can elucidate a new dimension in cancer pathogenesis. As such, miRNAs exemplify new potential therapeutic targets with a significant effect on both tumor growth and metastatic potential.

Microvascular density as an independent predictor of clinical outcome in renal cell carcinoma: an automated image analysis study.

Tumor microvascular density (MVD) has been shown to correlate with the aggressiveness of several cancers. With the introduction of targeted anti-angiogenic therapy, assessment of MVD has the potential not only as a prognostic but also as a therapeutic marker. The significance of tumor vascularity in clear cell renal cell carcinoma (ccRCC) has been debated, with studies showing contradictory results. Previous studies were limited by manual quantification of MVD within a small area of tumor. Since then, the validity of this method has been questioned. To avoid the inaccuracies of manual quantification, we employed a computerized image analysis, which allowed assessment of large areas of tumor and adjacent normal tissue. The latter was used as an internal reference for normalization. MVD and vascular endothelial growth factor (VEGF) were assessed in 57 cases of ccRCC. Sections were immunostained for CD34 and VEGF. Areas of ccRCC and normal kidney medulla were analyzed within scanned images using software that counted CD34-positive vessels and measured the intensity of VEGF staining. We obtained unadjusted values from tumoral areas and calculated adjusted values as tumor/normal ratios. Unadjusted MVD had no association with clinical outcome. However, similarly to tumor stage, higher adjusted MVD was associated with shorter disease-free survival (log-rank P=0.037, Cox P=0.02). This was significant in univariate and multivariate analyses. MVD did not correlate with tumor stage, pointing to its independent prognostic value. As expected due to the known molecular abnormalities in ccRCC, most tumors showed higher VEGF expression than normal tissue. Higher adjusted VEGF was associated with high tumor grade (P=0.049). The finding of increased MVD as an independent marker of tumor aggressiveness may prove useful in the development of new tests for prognostic and therapeutic guidance. Digital techniques can provide more accurate assessment of immunomarkers and may reveal less obvious associations.

Assessment of the prognostic significance of endoglin (CD105) in clear cell renal cell carcinoma using automated image analysis.

The behavior of clear cell renal cell carcinoma can be difficult to predict. Angiogenesis has proven to be a useful prognostic indicator in different malignancies. Endoglin (CD105) is a new marker of angiogenesis found to have prognostic utility in various tumors. Here, we provide the first automated digital assessment of intratumoral microvascular density in clear cell renal cell carcinoma using endoglin and CD31 and assess their utility as predictors of clinical outcome. Both endoglin and CD31 expression showed association with advanced tumor stage (P = .025 and P = .011, respectively). There was a significant correlation between CD31 and tumor grade (P = .034). Kaplan-Meier survival curves showed that patients with higher endoglin expression had significantly shorter progression-free survival (P = .010). Patients with higher CD31 expression tended to have a worse prognosis, although this was not statistically significant (P = .082). In univariate analysis using endoglin as a continuous variable, increased endoglin was strongly associated with reduced survival (hazard ratio, 1.74; 95% CI, 1.39-2.18; P = <.001). CD31 also correlated with poor outcomes (hazard ratio, 1.52; 95% CI, 1.24-1.86; P = .001). There was no correlation between CD31 and endoglin expression (r = -0.090, P = .541). Receiver operating characteristic analysis showed the area under the curve to be 0.749 for endoglin and 0.550 for CD31. In conclusion, increased endoglin and CD31 expression are associated with a higher tumor stage and decreased progression-free survival. Our automated approach overcomes many limitations of manual quantification. Advances in digital assessment of immunohistochemical markers can be helpful in standardizing the evaluation of tumor biomarkers.

miRNA profiling for clear cell renal cell carcinoma: biomarker discovery and identification of potential controls and consequences of miRNA dysregulation.

PURPOSE: Renal cell carcinoma is the most common neoplasm of the adult kidney. Currently to our knowledge there are no biomarkers for diagnostic, prognostic or predictive applications for renal cell carcinoma. miRNAs are nonprotein coding RNAs that negatively regulate gene expression and are potential biomarkers for cancer. MATERIALS AND METHODS: We analyzed 70 matched pairs of clear cell renal cell carcinoma and normal kidney tissues from the same patients by microarray analysis and validated our results by quantitative real-time polymerase chain reaction. We also performed extensive bioinformatic analysis to explore the role and regulation of miRNAs in clear cell renal cell carcinoma. RESULTS: We identified 166 miRNAs that were significantly dysregulated in clear cell renal cell carcinoma, including miR-122, miR-155 and miR-210, which had the highest over expression, and miR-200c, miR-335 and miR-218, which were most down-regulated. Analysis of previously reported miRNAs dysregulated in RCC showed overall agreement in the direction of dysregulation. Extensive target prediction analysis revealed that many miRNAs were predicted to target genes involved in renal cell carcinoma pathogenesis. In renal cell carcinoma miRNA dysregulation can be attributed in part to chromosomal aberrations, co-regulation of miRNA clusters and co-expression with host genes. We also performed a preliminary analysis showing that miR-155 expression correlated with clear cell renal cell carcinoma size. This finding must be validated in a larger independent cohort. CONCLUSIONS: Analysis showed that miRNAs are dysregulated in clear cell renal cell carcinoma and may contribute to kidney cancer pathogenesis by targeting more than 1 key molecule. We identified mechanisms that may contribute to miRNA dysregulation in clear cell renal cell carcinoma. Dysregulated miRNAs represent potential biomarkers for kidney cancer.

The miR-17-92 cluster is over expressed in and has an oncogenic effect on renal cell carcinoma.

PURPOSE: miRNAs are small, nonprotein coding RNAs that are differentially expressed in many malignancies. We previously identified 80 miRNAs that are dysregulated in clear cell renal cell carcinoma. In this study we validated over expression of the miR-17-92 cluster in clear cell renal cell carcinoma and tested the effect of 2 members of this cluster (miR-17-5p and miR-20a) on tumor proliferation. We also elucidated the role of miRNA in clear cell renal cell carcinoma pathogenesis with bioinformatics. MATERIALS AND METHODS: miRNA expression was validated by quantitative reverse transcriptase-polymerase chain reaction. The cell proliferation effect of miR-17-5p and miR-20a was tested in a renal adenocarcinoma cell line model. Multiple in silico analyses were done of dysregulated miRNAs. RESULTS: We validated miR-71-92 cluster over expression in clear cell renal cell carcinoma by quantitative reverse transcriptase-polymerase chain reaction. Transfection of miR-20a inhibitor significantly decreased cell proliferation in a dose dependent manner. Transfection of miR-17-5p, which is not endogenously expressed in the ACHN cell line, led to increased cell proliferation compared to control values. This effect was suppressed by miR-17-5p inhibitor. Bioinformatics analysis identified 10 clusters of miRNAs dysregulated in clear cell renal cell carcinoma that followed the same expression patterns. We also identified matching patterns between reported chromosomal aberration in clear cell renal cell carcinoma and miRNA dysregulation for 37.5% of the miRNAs. Target prediction analysis was done using multiple algorithms. Many key molecules in clear cell renal cell carcinoma pathogenesis, including HIFs, mTOR, VEGF and VHL, were potential targets for dysregulated miRNAs. CONCLUSIONS: A significant number of dysregulated proteins in clear cell renal cell carcinoma are potential miRNA targets. Also, many clear cell renal cell carcinoma dysregulated miRNAs are phylogenetically conserved.