ABSTRACT
OBJECTIVE: Recent evidence suggests that the fimbriated end of the fallopian tube harbors the precursor cells for many high-grade ovarian cancers, opening the door for development of better screening methods that directly assess the fallopian tube in women at risk for malignancy. Previously we have shown that the karyometric signature is abnormal in the fallopian tube epithelium in women at hereditary risk of ovarian cancer. In this study, we sought to determine whether the karyometric signature in serous tubal intraepithelial carcinoma (STIC) is significantly different from normal, and whether an abnormal karyometric signature can be detected in histologically normal tubal epithelial cells adjacent to STIC lesions. METHODS: The karyometric signature was measured in epithelial cells from the proximal and fimbriated portion of the fallopian tube in fallopian tube specimens removed from women at: 1) average risk for ovarian cancer undergoing surgery for benign gynecologic indications (n = 37), 2) hereditary risk of ovarian cancer (germline BRCA alterations) undergoing risk-reducing surgery (n = 44), and 3) diagnosed with fimbrial STICs (n = 17). RESULTS: The karyometric signature in tubes with fimbrial STICs differed from that of tubes with benign histology. The degree of karyometric alteration increased with increasing proximity to fimbrial STICs, ranging from moderate in the proximal portion of the tube, to greatest in both normal appearing fimbrial cells near STICs as well as in fimbrial STIC lesions. CONCLUSION: These data demonstrate an abnormal karyometric signature in STICs that may extend beyond the STIC, potentially providing an opportunity for early detection of fallopian tube neoplasia.
ABSTRACT
BACKGROUND: Many studies that aim to identify gene biomarkers using statistical methods and translate them into FDA-approved drugs have faced challenges due to lack of clinical validity and methodological reproducibility. Since genomic data analysis relies heavily on these statistical learning tools more than before, it is vital to address the limitations of these computational techniques. METHODS: Our study demonstrates these methodological gaps among most common statistical learning techniques used in gene expression analysis. To assess the classification ability and reproducibility of statistical learning tools for gene biomarker detection, six state-of-the-art machine learning models were trained on four different cancer data retrieved from The Cancer Genome Atlas (TCGA). Standard performance metrics including specificity, sensitivity, precision, and F1 score were evaluated to investigate the classification ability. For analysis of reproducibility, the identifiability of gene classifiers was examined by quantifying the consistency of the chosen classifier genes. RESULTS: Among the six state-of-the-art machine learning methods, the random forest had the best classification ability overall. Very few genes were selected by multiple methods, which suggests poor identifiability and reproducibility of statistical learning methods for gene expression data. Our results demonstrated the challenges of reproducing discoveries from gene expression analysis due to the inherent differences that exist in statistical machine learning methods. CONCLUSION: Since statistical machine learning models can have large variations in high-dimensional settings such as analysis of gene expression data, transparent analysis procedures including data preprocessing, model parameterization, and evaluation and choice of interpretable models are required for clinical validity and utility.
