Genomic determinants of prognosis in colorectal cancer

Applications of Machine Learning in Cancer Prediction and Prognosis

Machine learning is a branch of artificial intelligence that employs a variety of statistical, probabilistic and optimization techniques that allows computers to “learn” from past examples and to detect hard-to-discern patterns from large, noisy or complex data sets. This capability is particularly well-suited to medical applications, especially those that depend on complex proteomic and genomic measurements. As a result, machine learning is frequently used in cancer diagnosis and detection. More recently machine learning has been applied to cancer prognosis and prediction. This latter approach is particularly interesting as it is part of a growing trend towards personalized, predictive medicine. In assembling this review we conducted a broad survey of the different types of machine learning methods being used, the types of data being integrated and the performance of these methods in cancer prediction and prognosis. A number of trends are noted, including a growing dependence on protein biomarkers and microarray data, a strong bias towards applications in prostate and breast cancer, and a heavy reliance on “older” technologies such artificial neural networks (ANNs) instead of more recently developed or more easily interpretable machine learning methods. A number of published studies also appear to lack an appropriate level of validation or testing. Among the better designed and validated studies it is clear that machine learning methods can be used to substantially (15–25%) improve the accuracy of predicting cancer susceptibility, recurrence and mortality. At a more fundamental level, it is also evident that machine learning is also helping to improve our basic understanding of cancer development and progression.

A complex microsatellite at chromosome 7q33 as a new prognostic marker of colorectal cancer

Disease-specific markers are critical for early diagnosis, targeted therapy and prognostic prediction of diseases. Current study reports a complex microsatellite as a new prognostic marker of sporadic colorectal cancer. This microsatellite located at Chromosome 7q33 is composed of three tetranucleotide tandem repeats, (TTCC)₂(TCCC)₅(TCCT)₇, flanked by a CT-rich sequence. We analyzed polymorphisms of this microsatellite in 158 sporadic colorectal cancer, 143 matched normal adjacent tissues (NAT) and 150 health donors. Our results showed that this complex microsatellite was instable with polymorphic frequency of 77.2% in colorectal cancer, 52.4% in NAT and 54.7% in health donors (p<0.01) when compared to reference sequence. In the three tandem repeats, (TCCT)₇ site was most polymorphic accounting for over 70.0% of polymorphisms in this complex microsatellite, followed by (TTCC)₂ site for approximately 20%. Polymorphisms in (TCCC)₅ was rare. Polymorphisms at the (TCCT)₇ site were mainly insertions of 1 to 4 copies of TCCT (88.6%), and deletions occurred in about 6.4% of cases. The (TTCC)₂ site was featured with one copy TTCC insertions. Pair-wise analyses between colorectal tumors and NAT revealed that 88 of 121 (72.7%) tumors displayed expansion, contraction or both in these tetranucleotide tandem repeats when compared to NAT. A cross-analysis with clinicopathological data of 158 colorectal cancers revealed that polymorphic alterations of the microsatellite associated with less lymphatic metastasis (p<0.001), and the colorectal cancer patients with polymorphic changes in this microsatellite demonstrated better survival (n=112, p=0.0058). Together these data suggest that this complex microsatellite is a potential prognostic marker of sporadic colorectal cancer.

The follow-up after radical surgery of colorectal cancer: is it time for a "tailored" strategy?

The problem of the surveillance for colorectal cancer after radical surgery is a widely debated argument. Like for other solid tumors, the issue is divided in 2 main routes: the early diagnosis of recurrence and the early diagnosis of a second primary cancer. Genetic and molecular features have been recognized as useful tools to measure these risks, however, the instruments are still insufficient to design a personalized strategy for the patient. In an era of "tailored therapies" in oncology, even the follow-up of the surgically treated patient for colorectal cancer should enter "a tailor's shop in which several competent tailors" should be available to manage a complex problem.

Mismatch repair, p53 and chromosomal aberrations in primary colorectal carcinomas

Colorectal carcinoma progresses via at least two genetic pathways. Microsatellite instability, due to defective mismatch repair genes, characterizes one pathway and gross chromosomal instability another. The involvement of p53 and mismatch repair gene abnormalities within these pathways has not been fully explored. We aimed to investigate the relationships of p53 and mismatch repair gene defects on gross chromosomal aberrations detected by comparative genomic hybridization in 49 colorectal carcinomas. Tumours demonstrating loss of expression for hMLH1 or hMSH2 proteins demonstrated a highly significant attenuation in the number of gross chromosomal aberrations (p = 0.007) and were less likely to show p53 overexpression (p = 0.02). Within the mismatch repair normal tumours, p53 status did not affect the total number of chromosomal aberrations but p53 overexpression was significantly associated with a higher frequency of amplifications at 8q22-ter and at 13q21-22. Colorectal cancer demonstrates distinct molecular phenotypes and should be sub-classified accordingly.

Long-range epigenetic silencing at 2q14.2 affects most human colorectal cancers and may have application as a non-invasive biomarker of disease

Large chromosomal regions can be suppressed in cancer cells as denoted by hypermethylation of neighbouring CpG islands and downregulation of most genes within the region. We have analysed the extent and prevalence of long-range epigenetic silencing at 2q14.2 (the first and best characterised example of coordinated epigenetic remodelling) and investigated its possible applicability as a non-invasive diagnostic marker of human colorectal cancer using different approaches and biological samples. Hypermethylation of at least one of the CpG islands analysed (EN1, SCTR, INHBB) occurred in most carcinomas (90%), with EN1 methylated in 73 and 40% of carcinomas and adenomas, respectively. Gene suppression was a common phenomenon in all the tumours analysed and affected both methylated and unmethylated genes. Detection of methylated EN1 using bisulfite treatment and melting curve (MC) analysis from stool DNA in patients and controls resulted in a predictive capacity of, 44% sensitivity in positive patients (27% of overall sensitivity) and 97% specificity. We conclude that epigenetic suppression along 2q14.2 is common to most colorectal cancers and the presence of a methylated EN1 CpG island in stool DNA might be used as biomarker of neoplastic disease.

Microarray analysis of gene expression of mouse hepatocytes of different ploidy

Polyploidisation in hepatocytes has been associated with many physiologic and pathologic processes such as proliferation, metabolism, regeneration, aging, and cancer. We studied gene expression patterns in hepatocytes of different ploidy. Primary hepatocytes were obtained from mice of different ages: young (4-6 weeks old), adult (8-10 weeks old), and older (22-24 weeks old). Diploid (2N), tetraploid (4N), and octoploid (8N) hepatocytes were isolated for studies using a high-density mouse genome microarray. No major changes of gene expression patterns between hepatocytes of different ploidy were found. Fifty genes were identified as differentially expressed in the diploid and tetraploid populations, but the changes were less than twofold either way. Four genes (Gas2, Igfbp2, Nr1i3, and Ccne2) were differentially expressed in tetraploid and octoploid cells. This was confirmed in two age groups, "adult" and "older," but once again the factors were less than twofold and the expressions of Gas2 and Igfbp2 were more different between age groups than between ploidy classes. Our results show that polyploid hepatocytes are stable and "normal" without aberrant gene expression, unlike what is thought for cancer cells. By contrast to megakaryocytes, hepatocyte polyploidisation is not a differentiation step associated with major changes in gene expression. Our data support the hypothesis that hepatocyte polyploidisation is a protective mechanism against oxidative stress that occurs via a controlled process throughout growth and aging where binucleation is important.

Telomeres: prognostic markers for solid tumors

Solid tumors continue to affect millions of people worldwide. Increasingly sophisticated diagnostic tools contribute to the high incidence rates for some tumor types, and treatment options continue to expand. However, the progression of solid tumors represents a challenge for the appropriate treatment of individual patients because of the relative inaccuracy of current prognostic markers, including the widely used Tumor-Nodes-Metastasis (TNM) staging system, to predict the course of disease. As a result, both over- and undertreatment are clinical realities in the management of patients diagnosed with solid tumors. Therefore, population-based screening programs that increase the overall cancer incidence rates are controversial, as they may do little to improve the patient's quality of life. Consequently, there is a strong need to develop novel and independent markers of prognosis. In this context, we review the use of telomeres as prognostic markers for solid tumors, including cancers from lung, breast, prostate, colon, brain and head and neck. Telomeric sequences, the repetitive DNA at the end of human chromosomes, are mediators of genomic stability and can undergo length alterations during tumor initiation and progression. In a number of studies reviewed here, these alterations, measured as telomere attrition and elongation, have been shown either to be associated with clinical markers of disease progression or to be independent markers of cancer prognosis. We conclude from these studies that careful assessment of telomere length or its proxies, such as telomere DNA content, will be part of novel risk assessment and prognostic modalities for patients with solid tumors.

Clinical Implications of Our Advancing Knowledge of Colorectal Cancer Genetics: Inherited Syndromes, Prognosis, Prevention, Screening and Therapeutics

Recent genetic advances in our knowledge of colorectal cancer genetics are beginning to pay translational dividends in the management of this common clinical problem. We are now able to accurately screen and counsel individuals at risk of rare inherited cancer syndromes. We have recently introduced two of what are sure to be numerous biologic-based therapies, and have shown that colorectal neoplasia risk can be modestly reduced by various chemopreventative agents. Finally, our advancing knowledge has led to significant inroads into understanding what genetic alterations define prognosis and predict response to specific chemotherapeutic agents, and we are beginning to explore the utility of this knowledge in mass genetic-based clinical screening efforts. Enthusiasm must be tempered, however, by the extraordinary cost that often accompanies relatively modest gains. Finally, although genetic-based therapy often receives the greatest attention, molecular genetics, will likely have the greatest cost-effective impact in primary prevention and early diagnosis.

Mutant KRAS, chromosomal instability and prognosis in colorectal cancer

The RAS gene family provides a global effect on gene expression by encoding small GTP-binding proteins which act as molecular switches connecting extracellular signals with nuclear transcription factors. While wild type RAS proteins are switched off shortly after activation, mutant RAS proteins remain constitutively activated leading to complex interactions among their downstream effectors. For some human tumor types, these interactions were shown to contribute to cancer genesis and progression by inducing changes in cell survival, apoptosis, angiogenesis, invasion and metastasis. This review addresses the controversial link of KRAS mutations in colorectal cancer with chromosomal instability and patient prognosis.