Empathetic application of machine learning may address appropriate utilization of ART.

The value of artificial intelligence to benefit infertile patients is a subject of debate. This paper presents the experience of one aspect of artificial intelligence, machine learning, coupled with patient empathy to improve utilization of assisted reproductive technology (ART), which is an important aspect of care that is under-recognized. Although ART provides very effective options for infertile patients to build families, patients often discontinue ART when further treatment is likely to be beneficial and most of these patients do not achieve pregnancy without medical aid. Use of ART is only in part dependent on financial considerations; stress and other factors play a major role, as shown by high discontinuation rates despite reimbursement. This commentary discusses challenges and strategies to providing personalized ART prognostics based on machine learning, and presents a case study where appropriate use of such prognostics in ART centres is associated with a trend towards increased ART utilization.

Antimüllerian hormone levels and antral follicle count as prognostic indicators in a personalized prediction model of live birth.

OBJECTIVE: To compare antimüllerian hormone (AMH) and antral follicle count (AFC) separately and in combination with clinical characteristics for the prediction of live birth after controlled ovarian stimulation. DESIGN: Retrospective development and temporal external validation of prediction model. SETTING: Outpatient IVF clinic. PATIENT(S): We applied the boosted tree method to develop three prediction models incorporating clinical characteristics plus AMH or AFC or the combination on 2,124 linked IVF cycles from 2006 to 2010 and temporally externally validated predicted live-birth probabilities with an independent data set comprising 1,121 cycles from 2011 to 2012. INTERVENTION(S): None. MAIN OUTCOME MEASURE(S): Predictive power (posterior log of odds ratio compared to age, or PLORA), reclassification, receiver operator characteristic analysis, calibration, dynamic range. RESULT(S): Predictive power, was highest for the AMH model (PLORA = 29.1), followed by the AMH-AFC model (PLORA = 28.3) and AFC model (PLORA = 22.5). The prediction errors were 1% to <5% in each prognostic tier for all three models, except for the predicted live-birth probabilities of <10% in the AFC model, where the prediction error was 8%. The improvement in predictive power was highest for the AMH model: 76.2% improvement over age alone relative to 59% improvement for AFC and 73.3% for the combined model. Receiver operating characteristic analysis demonstrated that the AMH and the combined model had comparable discrimination (area under the curve = 0.716) and similar prediction error for high and low strata of live-birth prediction, with an improvement of 6.3% over age alone. CONCLUSION(S): The validated prediction model confirmed that AMH when combined with clinical characteristics can accurately identify the likelihood of live birth with a low prediction error. AFC provided no added predictive value beyond AMH.

Personalized prediction of first-cycle in vitro fertilization success.

OBJECTIVE: To test whether the probability of having a live birth (LB) with the first IVF cycle (C1) can be predicted and personalized for patients in diverse environments. DESIGN: Retrospective validation of multicenter prediction model. SETTING: Three university-affiliated outpatient IVF clinics located in different countries. PATIENT(S): Using primary models aggregated from >13,000 C1s, we applied the boosted tree method to train a preIVF-diversity model (PreIVF-D) with 1,061 C1s from 2008 to 2009, and validated predicted LB probabilities with an independent dataset comprising 1,058 C1s from 2008 to 2009. INTERVENTION(S): None. MAIN OUTCOME MEASURE(S): Predictive power, reclassification, receiver operator characteristic analysis, calibration, dynamic range. RESULT(S): Overall, with PreIVF-D, 86% of cases had significantly different LB probabilities compared with age control, and more than one-half had higher LB probabilities. Specifically, 42% of patients could have been identified by PreIVF-D to have a personalized predicted success rate >45%, whereas an age-control model could not differentiate them from others. Furthermore, PreIVF-D showed improved predictive power, with 36% improved log-likelihood (or 9.0-fold by log-scale; >1,000-fold linear scale), and prediction errors for subgroups ranged from 0.9% to 3.7%. CONCLUSION(S): Validated prediction of personalized LB probabilities from diverse multiple sources identify excellent prognoses in more than one-half of patients.

The chromatin remodeling factor Chd1l is required in the preimplantation embryo.

During preimplantation development, the embryo must establish totipotency and enact the earliest differentiation choices, processes that involve extensive chromatin modification. To identify novel developmental regulators, we screened for genes that are preferentially transcribed in the pluripotent inner cell mass (ICM) of the mouse blastocyst. Genes that encode chromatin remodeling factors were prominently represented in the ICM, including Chd1l, a member of the Snf2 gene family. Chd1l is developmentally regulated and expressed in embryonic stem (ES) cells, but its role in development has not been investigated. Here we show that inhibiting Chd1l protein production by microinjection of antisense morpholinos causes arrest prior to the blastocyst stage. Despite this important function in vivo, Chd1l is non-essential for cultured ES cell survival, pluripotency, or differentiation, suggesting that Chd1l is vital for events in embryos that are distinct from events in ES cells. Our data reveal a novel role for the chromatin remodeling factor Chd1l in the earliest cell divisions of mammalian development.

An Oct4-Sall4-Nanog network controls developmental progression in the pre-implantation mouse embryo.

Landmark events occur in a coordinated manner during pre-implantation development of the mammalian embryo, yet the regulatory network that orchestrates these events remains largely unknown. Here, we present the first systematic investigation of the network in pre-implantation mouse embryos using morpholino-mediated gene knockdowns of key embryonic stem cell (ESC) factors followed by detailed transcriptome analysis of pooled embryos, single embryos, and individual blastomeres. We delineated the regulons of Oct4, Sall4, and Nanog and identified a set of metabolism- and transport-related genes that were controlled by these transcription factors in embryos but not in ESCs. Strikingly, the knockdown embryos arrested at a range of developmental stages. We provided evidence that the DNA methyltransferase Dnmt3b has a role in determining the extent to which a knockdown embryo can develop. We further showed that the feed-forward loop comprising Dnmt3b, the pluripotency factors, and the miR-290-295 cluster exemplifies a network motif that buffers embryos against gene expression noise. Our findings indicate that Oct4, Sall4, and Nanog form a robust and integrated network to govern mammalian pre-implantation development.

Predicting personalized multiple birth risks after in vitro fertilization-double embryo transfer.

OBJECTIVE: To report and evaluate the performance and utility of an approach to predicting IVF-double embryo transfer (DET) multiple birth risks that is evidence-based, clinic-specific, and considers each patient's clinical profile. DESIGN: Retrospective prediction modeling. SETTING: An outpatient university-affiliated IVF clinic. PATIENT(S): We used boosted tree methods to analyze 2,413 independent IVF-DET treatment cycles that resulted in live births. The IVF cycles were retrieved from a database that comprised more than 33,000 IVF cycles. INTERVENTION(S): None. MAIN OUTCOME MEASURE(S): The performance of this prediction model, MBP-BIVF, was validated by an independent data set, to evaluate predictive power, discrimination, dynamic range, and reclassification. RESULT(S): Multiple birth probabilities ranging from 11.8% to 54.8% were predicted by the model and were significantly different from control predictions in more than half of the patients. The prediction model showed an improvement of 146% in predictive power and 16.0% in discrimination over control. The population standard error was 1.8%. CONCLUSION(S): We showed that IVF patients have inherently different risks of multiple birth, even when DET is specified, and this risk can be predicted before ET. The use of clinic-specific prediction models provides an evidence-based and personalized method to counsel patients.

Single cell transcriptional profiling reveals heterogeneity of human induced pluripotent stem cells.

Human induced pluripotent stem cells (hiPSCs) and human embryonic stem cells (hESCs) are promising candidate cell sources for regenerative medicine. However, despite the common ability of hiPSCs and hESCs to differentiate into all 3 germ layers, their functional equivalence at the single cell level remains to be demonstrated. Moreover, single cell heterogeneity amongst stem cell populations may underlie important cell fate decisions. Here, we used single cell analysis to resolve the gene expression profiles of 362 hiPSCs and hESCs for an array of 42 genes that characterize the pluripotent and differentiated states. Comparison between single hESCs and single hiPSCs revealed markedly more heterogeneity in gene expression levels in the hiPSCs, suggesting that hiPSCs occupy an alternate, less stable pluripotent state. hiPSCs also displayed slower growth kinetics and impaired directed differentiation as compared with hESCs. Our results suggest that caution should be exercised before assuming that hiPSCs occupy a pluripotent state equivalent to that of hESCs, particularly when producing differentiated cells for regenerative medicine aims.

Deep phenotyping to predict live birth outcomes in in vitro fertilization.

Nearly 75% of in vitro fertilization (IVF) treatments do not result in live births and patients are largely guided by a generalized age-based prognostic stratification. We sought to provide personalized and validated prognosis by using available clinical and embryo data from prior, failed treatments to predict live birth probabilities in the subsequent treatment. We generated a boosted tree model, IVFBT, by training it with IVF outcomes data from 1,676 first cycles (C1s) from 2003-2006, followed by external validation with 634 cycles from 2007-2008, respectively. We tested whether this model could predict the probability of having a live birth in the subsequent treatment (C2). By using nondeterministic methods to identify prognostic factors and their relative nonredundant contribution, we generated a prediction model, IVF(BT), that was superior to the age-based control by providing over 1,000-fold improvement to fit new data (p<0.05), and increased discrimination by receiver-operative characteristic analysis (area-under-the-curve, 0.80 vs. 0.68 for C1, 0.68 vs. 0.58 for C2). IVFBT provided predictions that were more accurate for approximately 83% of C1 and approximately 60% of C2 cycles that were out of the range predicted by age. Over half of those patients were reclassified to have higher live birth probabilities. We showed that data from a prior cycle could be used effectively to provide personalized and validated live birth probabilities in a subsequent cycle. Our approach may be replicated and further validated in other IVF clinics.

Maternally and zygotically provided Cdx2 have novel and critical roles for early development of the mouse embryo.

Divisions of polarised blastomeres that allocate polar cells to outer and apolar cells to inner positions initiate the first cell fate decision in the mouse embryo. Subsequently, outer cells differentiate into trophectoderm while inner cells retain pluripotency to become inner cell mass (ICM) of the blastocyst. Elimination of zygotic expression of trophectoderm-specific transcription factor Cdx2 leads to defects in the maintenance of the blastocyst cavity, suggesting that it participates only in the late stage of trophectoderm formation. However, we now find that mouse embryos also have a maternally provided pool of Cdx2 mRNA. Moreover, depletion of both maternal and zygotic Cdx2 from immediately after fertilization by three independent approaches, dsRNAi, siRNAi and morpholino oligonucleotides, leads to developmental arrest at much earlier stages than expected from elimination of only zygotic Cdx2. This developmental arrest is associated with defects in cell polarisation, reflected by expression and localisation of cell polarity molecules such as Par3 and aPKC and cell compaction at the 8- and 16-cell stages. Cells deprived of Cdx2 show delayed development with increased cell cycle length, irregular cell division and increased incidence of apoptosis. Although some Cdx2-depleted embryos initiate cavitation, the cavity cannot be maintained. Furthermore, expression of trophectoderm-specific genes, Gata3 and Eomes, and also the trophectoderm-specific cytokeratin intermediate filament, recognised by Troma1, are greatly reduced or undetectable. Taken together, our results indicate that Cdx2 participates in two steps leading to trophectoderm specification: appropriate polarisation of blastomeres at the 8- and 16-cell stage and then the maintenance of trophectoderm lineage-specific differentiation.

Gene expression profiles of human inner cell mass cells and embryonic stem cells.

Human embryonic stem cell (hESC) lines are derived from the inner cell mass (ICM) of preimplantation human blastocysts obtained on days 5-6 following fertilization. Based on their derivation, they were once thought to be the equivalent of the ICM. Recently, however, studies in mice reported the derivation of mouse embryonic stem cell lines from the epiblast; these epiblast lines bear significant resemblance to human embryonic stem cell lines in terms of culture, differentiation potential and gene expression. In this study, we compared gene expression in human ICM cells isolated from the blastocyst and embryonic stem cells. We demonstrate that expression profiles of ICM clusters from single embryos and hESC populations were highly reproducible. Moreover, comparison of global gene expression between individual ICM clusters and human embryonic stem cells indicated that these two cell types are significantly different in regards to gene expression, with fewer than one half of all genes expressed in both cell types. Genes of the isolated human inner cell mass that are upregulated and downregulated are involved in numerous cellular pathways and processes; a subset of these genes may impart unique characteristics to hESCs such as proliferative and self-renewal properties.

In vitro embryo culture in defined, sub-microliter volumes.

The high attrition rate of in vitro human embryo culture presents a major obstacle in the treatment of clinical infertility by in vitro fertilization (IVF). Physical and genetic requirements are not well understood for human or mouse preimplantation embryo development. Group culture is an established requirement for optimal embryo development in the mouse model. However, conventional microdrop culture limitations hinder investigations of the effects of physical parameters on in vitro embryo development. We report a microfluidics platform that enables embryo culture in precisely defined, sub-microliter volumes (5-500 nl) which cannot be investigated using conventional methods. Groups of two embryos per microfluidic well successfully developed to the blastocyst stage, at a rate of over 80%, which is comparable to those cultured in 20-microl microdrops. This system can be used to dissect physical requirements of in vitro single or group embryo culture, and be made highly parallel to increase experimental throughput.

Defining human embryo phenotypes by cohort-specific prognostic factors.

BACKGROUND: Hundreds of thousands of human embryos are cultured yearly at in vitro fertilization (IVF) centers worldwide, yet the vast majority fail to develop in culture or following transfer to the uterus. However, human embryo phenotypes have not been formally defined, and current criteria for embryo transfer largely focus on characteristics of individual embryos. We hypothesized that embryo cohort-specific variables describing sibling embryos as a group may predict developmental competence as measured by IVF cycle outcomes and serve to define human embryo phenotypes. METHODOLOGY/PRINCIPAL FINDINGS: We retrieved data for all 1117 IVF cycles performed in 2005 at Stanford University Medical Center, and further analyzed clinical data from the 665 fresh IVF, non-donor cycles and their associated 4144 embryos. Thirty variables representing patient characteristics, clinical diagnoses, treatment protocol, and embryo parameters were analyzed in an unbiased manner by regression tree models, based on dichotomous pregnancy outcomes defined by positive serum beta-human chorionic gonadotropin (beta-hCG). IVF cycle outcomes were most accurately predicted at approximately 70% by four non-redundant, embryo cohort-specific variables that, remarkably, were more informative than any measures of individual, transferred embryos: Total number of embryos, number of 8-cell embryos, rate (percentage) of cleavage arrest in the cohort and day 3 follicle stimulating hormone (FSH) level. While three of these variables captured the effects of other significant variables, only the rate of cleavage arrest was independent of any known variables. CONCLUSIONS/SIGNIFICANCE: Our findings support defining human embryo phenotypes by non-redundant, prognostic variables that are specific to sibling embryos in a cohort.

A novel and critical role for Oct4 as a regulator of the maternal-embryonic transition.

BACKGROUND: Compared to the emerging embryonic stem cell (ESC) gene network, little is known about the dynamic gene network that directs reprogramming in the early embryo. We hypothesized that Oct4, an ESC pluripotency regulator that is also highly expressed at the 1- to 2-cell stages in embryos, may be a critical regulator of the earliest gene network in the embryo. METHODOLOGY/PRINCIPAL FINDINGS: Using antisense morpholino oligonucleotide (MO)-mediated gene knockdown, we show that Oct4 is required for development prior to the blastocyst stage. Specifically, Oct4 has a novel and critical role in regulating genes that encode transcriptional and post-transcriptional regulators as early as the 2-cell stage. Our data suggest that the key function of Oct4 may be to switch the developmental program from one that is predominantly regulated by post-transcriptional control to one that depends on the transcriptional network. Further, we propose to rank candidate genes quantitatively based on the inter-embryo variation in their differential expression in response to Oct4 knockdown. Of over 30 genes analyzed according to this proposed paradigm, Rest and Mta2, both of which have established pluripotency functions in ESCs, were found to be the most tightly regulated by Oct4 at the 2-cell stage. CONCLUSIONS/SIGNIFICANCE: We show that the Oct4-regulated gene set at the 1- to 2-cell stages of early embryo development is large and distinct from its established network in ESCs. Further, our experimental approach can be applied to dissect the gene regulatory network of Oct4 and other pluripotency regulators to deconstruct the dynamic developmental program in the early embryo.

A transgenic mouse model for high content, cell cycle phenotype screening in live primary cells.

High content cell-based genetic and small molecule library screens are powerful strategies in drug discovery and investigations of disease mechanisms. We report that primary cells derived from a transgenic mouse model expressing a fluorescence mitosis biosensor provide unambiguous phenotype readouts without the need for transfection or immunocytochemistry. Phenotype profiles of cell cycle disruption and of apoptosis are easily detectable at a single time point selected from time-lapse live fluorescence microscopy. Most importantly, this transgenic mouse model may be crossed with cancer mouse models to derive biosensor-expressing primary cancer cells for use in high content screening strategies targeting discovery of tumor-specific chemotherapeutic compounds.

Gene expression profiling reveals progesterone-mediated cell cycle and immunoregulatory roles of Hoxa-10 in the preimplantation uterus.

Human infertility and recurrent pregnancy loss caused by implantation defects are poorly understood. Hoxa-10-deficient female mice have severe infertility and recurrent pregnancy loss due to defective uterine implantation. Gene expression profiling experiments reveal that Hoxa-10 is an important regulator of two critical events in implantation: stromal cell proliferation and local immunosuppression. At the time of implantation, Hoxa-10 mediates the progesterone-stimulated proliferation of uterine stromal cells. Hoxa-10 mutants express a stromal cell proliferation defect that is accompanied by quantitative or spatial alterations in the expression of two cyclin-dependent kinase inhibitor genes, p57 and p15. Hoxa-10 deficiency also leads to a severe local immunological disturbance, characterized by a polyclonal proliferation of T cells, that occurs in place of the normal progesterone-mediated immunosuppression in the periimplantation uterus.