The endometrium is a dynamic tissue that changes under the influence of hormones in order to create optimal conditions for embryo implantation. To better understand the molecular mechanisms of endometrial receptivity, several research groups have performed transcriptomic analysis of the endometrium of mice [57, 58], rats [59], and rhesus monkeys [60]. The human endometrium has been studied in pathological conditions (such as endometrial cancer) to better characterise the molecular pathways involved in pathogenesis [61] and throughout the normal menstrual cycle [62]. The latter study showed that the endometrium may be ‘dated’ to specific phase of menstrual cycle based on its transcriptional profile. Moreover, specific gene clusters characteristic of the different phases of the menstrual cycle have been described [62]. In IVF procedures, one of the major challenges has been to identify the endometrial window of receptivity and several groups have tried to find it by using transcriptomic analysis [63–66]. These investigators have generated extensive lists of genes proposed as markers of endometrial receptivity; however, only one gene—osteopontin—appears on all rosters. Osteopontin is involved in cell adhesion, but its role in human embryo implantation remains poorly understood [67].

Comparison of endometrial gene expression in unstimulated IVF cycles, stimulated IVF cycles, and immediately after removal of IUD showed there were 25 genes expressed during the window of implantation (WOI) in common for all three conditions [68]. Interestingly, these genes seemed to be regulated in one way in unstimulated cycles but in the opposite way in both stimulated and IUD cycles. In other words, if a gene was overexpressed in unstimulated cycles, it was downregulated in other two conditions and vice versa. Recently, the group of Simón [16] introduced an endometrial receptivity array (ERA) containing 238 genes, related to endometrial receptivity. By using ERA, we could determine an individual WOI for women with repeated implantation failure and thus perform the embryo transfer on an optimal day of IVF cycle.

To better understand the molecular mechanisms during preimplantation development, several studies have analysed global gene expression profile of embryos in humans [69–71] and mice [72, 73]. Precise control of gene expression during the preimplantation embryonic development is of particular significance. The first cellular differentiation occurs at this time and the embryo transfers from a reliance on maternal RNA derived from the oocyte to expression of its own genome. Wells et al. [70] examined the expression of nine known genes implicated in important cellular processes such as cell cycle regulation, DNA repair, apoptosis, maintenance of accurate chromosomal segregation, and construction of the cytoskeleton throughout the preimplantation phase of embryo development by using qPCR. The genes tested were BRCA1, BRCA2, ATM, TP53, RB1, MAD2, BUB1, APC, and β-actin. They established that the expression levels of all nine genes decreased dramatically after fertilisation and then recovered between the 4- and 8-cell embryo stages. Further increase of gene expression (or in some cases a slight reduction) was seen at the morula stage before gene expression jumped significantly at the blastocyst stage. Of note, global transcriptomic analysis of mouse embryos has revealed a requirement for maternal RNA depletion before embryonic genome activation. This process happens in two stages: zygotic genome activation and mid-preimplantation gene activation [72]. After zygotic genome activation de novo gene transcription begins, it is needed for morula to undergo morphological and functional changes and develop to blastocyst.

Many morphologically normal embryos do not achieve pregnancy after embryo transfer. It has been postulated that many failed IVF cycles occur because of chromosomally abnormal oocytes or embryos. For this reason, several groups have tried to screen oocytes and embryos for aneuploidies by using transcriptomics to identify euploid and viable oocytes and embryos with greatest chances for implantation [6, 53, 74]. Wells and Delhanty [75] introduced a molecular cytogenetic method allowing the simultaneous enumeration of all of the chromosomes in a single cell called comparative genomic hybridisation (CGH). They report an improvement in embryo implantation and pregnancy rates with the proportion of CGH screened embryos resulting in live birth was 80 % as compared to 60 % for patients without CGH screening [76, 77]. The major pitfalls of using CGH were the long time required for the method (approximately 4 days), which was incompatible with a fresh transfer timeframe during IVF.

With developing technology of transcriptomics, proteomics, and metabolomics, new biomarker identification has greatly accelerated. With that has come an intense discussion on how best to measure newly discovered biomarkers. Understandably, there is great interest of implementing discovered biomarkers into clinical practice. For example, the American Society of Clinical Oncology has presented a paper where it is estimated that routinely testing people with colon cancer would save at least US $600 million a year [78].

In the past decade, we have witnessed increased numbers of biomarker publications, but most of them do not have sufficient sensitivity and/or specificity to be clinically useful. This weakness is reflected in the relatively low number of patent applications and the even lower number of successful market applications [79] for such discoveries. The major pitfalls in the translation from biomarker discovery to clinical utility are:

In order to overcome these limitations, certain authorities [e.g. American Society of Clinical Oncology, U.S. Food and Drug Administration (FDA), European Medicines agency (EMA), European Association for Predictive, Preventive and Personalized Medicine (EPMA), National Institutes of Health (NIH)] have developed guidelines on validation process for studies of biomarker discovery. For the purpose of this chapter, these recommendations can be extracted by terms analytical validity, clinical validity, and clinical utility [80].

Biomarkers can be classified into the following categories: pharmacodynamic, prognostic, or predictive [81].

As a rule it can be considered that ideal biomarkers for use in diagnostics and prognostics, as well as for drug developing and targeting, should be highly specific and sensitive [82]. But in reality, only rare biomarkers have high sensitivity and specificity. According to Issaq et al. [82], the following factors attribute to this:

As described previously, many biomarkers have been proposed for various endpoints in IVF cycle (oocyte maturity, oocyte fertilisation, embryo quality, pregnancy). Biomarkers for pregnancy seem to be most appropriate for SET in clinical practice. But a major drawback of all biomarkers CC and GC thus far discovered remains the lack of validation. Only a few of all proposed biomarkers have been validated by any statistical method.

In the study of McKenzie et al., HAS2, PTGS2, and GREM1 were validated by a logistic regression model for oocyte maturity, oocyte fertilisation, and embryo quality. Regression models for embryo quality yielded an AUC 0.76, 0.76, and 0.81 for HAS2, PTGS2, and GREM1, respectively. Combining PTGS2 and GREM1 improved the predictive power only slightly (AUC 0.82 vs. 0.81) [38].

Whatlet et al. investigated PTGS2, SDC4, VCAN, GREM1, ITPKA, CALM2, and TRPM7 and used a multivariate regression model for embryo quality and pregnancy. Better cleavage-stage embryo prediction relied on TRPM7 and ITPKA expression, and the pregnancy prediction relied on SDC4 and VCAN expression. The developed multivariate regression models for prediction of pregnancy had a sensitivity of 0.70 and a specificity of 0.90 in the analysed dataset [44].