Machine learning within the Parkinson's progression markers initiative: Review of the current state of affairs.

The Parkinson's Progression Markers Initiative (PPMI) has collected more than a decade's worth of longitudinal and multi-modal data from patients, healthy controls, and at-risk individuals, including imaging, clinical, cognitive, and 'omics' biospecimens. Such a rich dataset presents unprecedented opportunities for biomarker discovery, patient subtyping, and prognostic prediction, but it also poses challenges that may require the development of novel methodological approaches to solve. In this review, we provide an overview of the application of machine learning methods to analyzing data from the PPMI cohort. We find that there is significant variability in the types of data, models, and validation procedures used across studies, and that much of what makes the PPMI data set unique (multi-modal and longitudinal observations) remains underutilized in most machine learning studies. We review each of these dimensions in detail and provide recommendations for future machine learning work using data from the PPMI cohort.

Optimal Intensity and Duration of Walking Rehabilitation in Patients With Chronic Stroke: A Randomized Clinical Trial.

Importance: For walking rehabilitation after stroke, training intensity and duration are critical dosing parameters that lack optimization. Objective: To assess the optimal training intensity (vigorous vs moderate) and minimum training duration (4, 8, or 12 weeks) needed to maximize immediate improvement in walking capacity in patients with chronic stroke. Design, Setting, and Participants: This multicenter randomized clinical trial using an intent-to-treat analysis was conducted from January 2019 to April 2022 at rehabilitation and exercise research laboratories. Survivors of a single stroke who were aged 40 to 80 years and had persistent walking limitations 6 months or more after the stroke were enrolled. Interventions: Participants were randomized 1:1 to high-intensity interval training (HIIT) or moderate-intensity aerobic training (MAT), each involving 45 minutes of walking practice 3 times per week for 12 weeks. The HIIT protocol used repeated 30-second bursts of walking at maximum safe speed, alternated with 30- to 60-second rest periods, targeting a mean aerobic intensity above 60% of the heart rate reserve (HRR). The MAT protocol used continuous walking with speed adjusted to maintain an initial target of 40% of the HRR, progressing up to 60% of the HRR as tolerated. Main Outcomes and Measures: The main outcome was 6-minute walk test distance. Outcomes were assessed by blinded raters after 4, 8, and 12 weeks of training. Results: Of 55 participants (mean [SD] age, 63 [10] years; 36 male [65.5%]), 27 were randomized to HIIT and 28 to MAT. The mean (SD) time since stroke was 2.5 (1.3) years, and mean (SD) 6-minute walk test distance at baseline was 239 (132) m. Participants attended 1675 of 1980 planned treatment visits (84.6%) and 197 of 220 planned testing visits (89.5%). No serious adverse events related to study procedures occurred. Groups had similar 6-minute walk test distance changes after 4 weeks (HIIT, 27 m [95% CI, 6-48 m]; MAT, 12 m [95% CI, -9 to 33 m]; mean difference, 15 m [95% CI, -13 to 42 m]; P = .28), but HIIT elicited greater gains after 8 weeks (58 m [95% CI, 39-76 m] vs 29 m [95% CI, 9-48 m]; mean difference, 29 m [95% CI, 5-54 m]; P = .02) and 12 weeks (71 m [95% CI, 49-94 m] vs 27 m [95% CI, 3-50 m]; mean difference, 44 m [95% CI, 14-74 m]; P = .005) of training; HIIT also showed greater improvements than MAT on some secondary measures of gait speed and fatigue. Conclusions and Relevance: These findings show proof of concept that vigorous training intensity is a critical dosing parameter for walking rehabilitation. In patients with chronic stroke, vigorous walking exercise produced significant and meaningful gains in walking capacity with only 4 weeks of training, but at least 12 weeks were needed to maximize immediate gains. Trial Registration: ClinicalTrials.gov Identifier: NCT03760016.

Fatty Acid Transfer from Blood to Milk Is Disrupted in Mothers with Low Milk Production, Obesity, and Inflammation.

BACKGROUND: Obesity is associated with chronic inflammation and is a risk factor for insufficient milk production. Inflammation-mediated suppression of LPL could inhibit mammary uptake of long-chain fatty acids (LCFAs; >16 carbons). OBJECTIVES: In an ancillary case-control analysis, we investigated whether women with low milk production despite regular breast emptying have elevated inflammation and disrupted transfer of LCFAs from plasma into milk. METHODS: Data and specimens from a low milk supply study and an exclusively breastfeeding control group were analyzed, with milk production measured by 24-h test-weighing at 2-10 wk postpartum. Low milk supply groups were defined as very low (VL; <300 mL/d; n = 23) or moderate (MOD; ≥300 mL/d; n = 20) milk production, and compared with controls (≥699 mL/d; n = 18). Serum and milk fatty acids (weight% of total) were measured by GC, serum and milk TNF-α by ELISA, and serum high-sensitivity C-reactive protein (hsCRP) by clinical analyzer. Group differences were assessed by linear regression models, chi-square exact tests, and Kruskal-Wallis nonparametric tests. RESULTS: VL cases, as compared with MOD cases and controls, had higher prevalence of elevated serum hsCRP (>5 mg/L; 57%, 15%, and 22%, respectively; P = 0.004), detectable milk TNF-α (67%, 32%, and 33%, respectively; P = 0.04), and obesity (78%, 40%, and 22%, respectively; P = 0.003). VL cases had lower mean ± SD LCFAs in milk (60% ± 3%) than MOD cases (65% ± 4%) and controls (66% ± 5%) (P < 0.001). Milk and serum LCFAs were strongly correlated in controls (r = 0.82, P < 0.001), but not in the MOD (r = 0.25, P = 0.30) or VL (r = 0.20, P = 0.41) groups (Pint < 0.001). CONCLUSIONS: Mothers with very low milk production have significantly higher obesity and inflammatory biomarkers, lower LCFAs in milk, and disrupted association between plasma and milk LCFAs. These data support the hypothesis that inflammation disrupts normal mammary gland fatty acid uptake. Further research should address impacts of inflammation and obesity on mammary fatty acid uptake for milk production.

Measures of Maternal Metabolic Health as Predictors of Severely Low Milk Production.

Background: A comprehensive approach to breastfeeding support requires elucidation of how metabolic health influences milk production. Objective: We compared metabolic health indicators in women with severely low milk output versus those with moderate/normal milk output using a case-control study design, with nested and external control groups. Design: Cases and nested controls were derived from women screened for a low milk supply trial, with cases defined as severely low milk output (<300 mL/24 hours), and nested controls defined as moderate/normal milk output (>300 mL/24 hours). In addition, we included an external control group of exclusively breastfeeding women. All were enrolled at 2-10 weeks postdelivery of a healthy term infant. Milk output and breast emptying frequency were recorded through test-weigh. Metabolic health variables included all components of the metabolic syndrome, homeostatic model assessment of insulin resistance (HOMA-IR), and diagnosis of gestational diabetes mellitus (GDM). Results: Maximum milk output, mL/24 hours, ranged as follows: 30-281 in cases (n = 18), 372-801 in nested controls (n = 12), and 661-915 in external controls (n = 12). Mean breast emptying frequency in cases was not significantly different from nested or external controls. All metabolic syndrome components and HOMA-IR were significantly worse in cases as compared with both nested and external control groups (p < 0.05). There was no significant difference between the nested and external control groups for these variables. GDM prevalence was 39%, 0%, and 8%, across cases, nested control, and external control groups, respectively (chi-square p-value = 0.02). Conclusion: Results from this small case-control study identify class 2+ obesity and poor metabolic health as strong risk factors for severely low milk production. These findings should be further validated in larger prospective cohort studies designed to identify individuals at risk for metabolically driven low milk supply. In addition, clinical and qualitative research studies aimed at improving patient-centered approaches to the management of persistent low milk supply are needed.

Moderate-intensity exercise versus high-intensity interval training to recover walking post-stroke: protocol for a randomized controlled trial.

BACKGROUND: Stroke results in neurologic impairments and aerobic deconditioning that contribute to limited walking capacity which is a major barrier post-stroke. Current exercise recommendations and stroke rehabilitation guidelines recommend moderate-intensity aerobic training post-stroke. Locomotor high-intensity interval training is a promising new strategy that has shown significantly greater improvements in aerobic fitness and motor performance than moderate-intensity aerobic training in other populations. However, the relative benefits and risks of high-intensity interval training and moderate-intensity aerobic training remain poorly understood following stroke. In this study, we hypothesize that locomotor high-intensity interval training will result in greater improvements in walking capacity than moderate-intensity aerobic training. METHODS: Using a single-blind, 3-site randomized controlled trial, 50 chronic (> 6 months) stroke survivors are randomly assigned to complete 36 locomotor training sessions of either high-intensity interval training or moderate-intensity aerobic training. Main eligibility criteria are age 40-80 years, single stroke for which the participant received treatment (experienced 6 months to 5 years prior to consent), walking speed ≤ 1.0 m/s, able to walk at least 3 min on the treadmill at ≥ 0.13 m/s (0.3 mph), stable cardiovascular condition (American Heart Association class B), and the ability to walk 10 m overground without continuous physical assistance. The primary outcome (walking capacity) and secondary outcomes (self-selected and fast gait speed, aerobic fitness, and fatigue) are assessed prior to initiating training and after 4 weeks, 8 weeks, and 12 weeks of training. DISCUSSION: This study will provide fundamental new knowledge to inform the selection of intensity and duration dosing parameters for gait recovery and optimization of aerobic training interventions in chronic stroke. Data needed to justify and design a subsequent definitive trial will also be obtained. Thus, the results of this study will inform future stroke rehabilitation guidelines on how to optimally improve walking capacity following stroke. TRIAL REGISTRATION: ClinicalTrials.gov NCT03760016 . Registered on November 30, 2018.

Validity of a 3-Hour Breast Milk Expression Protocol in Estimating Current Maternal Milk Production Capacity and Infant Breast Milk Intake in Exclusively Breastfeeding Dyads.

Objective: An efficient method for measuring maternal milk production is needed. Our objectives were to: (1) validate a milk production rate (MPR) protocol in exclusively breastfeeding mothers; (2) determine MPR change following 48 hours of increased breast emptying; (3) assess agreement between MPR and infant test-weighing; and (4) characterize MPR in early postpartum exclusively breastfeeding mothers. Materials and Methods:N = 23 mothers emptied both breasts hourly over 3 hours (h0, h1, h2, and h3). We estimated steady-state MPR as mean (h2 and h3). Subset A mothers (n = 5) also completed MPR measurements after 48 hours of increased breast emptying. Subset B mothers (n = 16) also test-weighed for 48 hours. We used paired t-test to examine within-participant change in hourly milk yield and MPR; and we used Bland-Altman analysis to compare 24-hour milk production (g/24 hours) measured using test-weight versus MPR. Results are reported as mean ± standard deviation or (±95% limits of agreement). Results: Mothers were 54 ± 14 days postpartum. Paired difference in h3-h2 hourly milk yield was not significantly different (p > 0.05, 3 ± 10 g/hour). In Subset A (n = 5), MPR declined from 50 ± 13 to 43 ± 16 g/hour (p = 0.003) following 48 hours of increased breast emptying. In Study B (n = 16), mean infant test-weighed intake (TW) was 717 ± 119 g/24 hours, and mean MPR was 1,085 ± 300 g/24 hours. Mean difference (MPR-test-weigh) and mean ratio (MPR/test-weigh) significantly increased as MPR increased (p < 0.05). For infants with adequate weight gain (>20 g/24 hours, n = 12), mean MPR = 48 ± 12 g/hour (range, 35-78 g/hour). Conclusion: MPR is a valid measure of current maternal milk production capacity, but is not accurate for evaluating infant intake in exclusively breastfeeding dyads.

Phase I Study of Ramucirumab Plus Merestinib in Previously Treated Metastatic Colorectal Cancer: Safety, Preliminary Efficacy, and Pharmacokinetic Findings.

LESSONS LEARNED: The combination of the antivascular endothelial growth factor receptor 2 monoclonal antibody, ramucirumab, and the type II MET kinase inhibitor, merestinib, is tolerable. Preliminary efficacy data suggest that the combination may provide clinical benefit to patients with metastatic colorectal cancer (mCRC). Further development of this combination would likely necessitate the identification of subsets of patients with mCRC where the clinical benefit is of clinical relevance. BACKGROUND: This study evaluated safety, preliminary efficacy, and pharmacokinetics of ramucirumab plus merestinib in patients with MCR previously treated with oxaliplatin and/or irinotecan. METHODS: Open-label phase Ia/b study comprising 3+3 dose-limiting toxicity (DLT) observation and expansion parts. Treatment was ramucirumab 8 mg/kg on days 1 and 15 and merestinib 80 mg once daily (QD; 28-day cycle). Primary objective was safety and tolerability. Secondary objectives were pharmacokinetics and preliminary antitumor activity. Exploratory objective was biomarker associations. RESULTS: Safety findings: DLT (proteinuria) of 7 phase Ia patients (the expansion part started at the initial recommended dose level); 16 patients (70%) with grade ≥3 treatment-emergent adverse events (TEAEs); 10 patients (43%) with grade ≥3 treatment-related TEAEs. The most common grade ≥3 treatment-related TEAEs were fatigue (4 patients [17%]) and increased blood alkaline phosphatase, diarrhea, and hypertension (2 patients each [9%]). One patient discontinued treatment because of cholestatic hepatitis. Geometric mean trough concentrations at cycle 1, day 15, were ramucirumab, 24.8 µg/mL; merestinib, 130 ng/mL. No complete or partial response was seen; 12 patients (52%) achieved stable disease. Median progression-free survival was 3.3 months (95% confidence interval [CI]: 1.6-4.4). Median overall survival was 8.9 months (95% CI: 3.5-12.7). There were no associations between genetic alterations and efficacy. CONCLUSION: Ramucirumab plus merestinib is tolerable and may have clinical benefit in biomarker-unselected, heavily pretreated patients with mCRC.

Open science: The open clinical trials data journey.

Open data sharing and access has the potential to promote transparency and reproducibility in research, contribute to education and training, and prompt innovative secondary research. Yet, there are many reasons why researchers don't share their data. These include, among others, time and resource constraints, patient data privacy issues, lack of access to appropriate funding, insufficient recognition of the data originators' contribution, and the concern that commercial or academic competitors may benefit from analyses based on shared data. Nevertheless, there is a positive interest within and across the research and patient communities to create shared data resources. In this perspective, we will try to highlight the spectrum of "openness" and "data access" that exists at present and highlight the strengths and weakness of current data access platforms, present current examples of data sharing platforms, and propose guidelines to revise current data sharing practices going forward.

Feasibility and Acceptability of Metformin to Augment Low Milk Supply: A Pilot Randomized Controlled Trial.

BACKGROUND: Metformin improves insulin action, but feasibility in treating low milk supply is unknown. RESEARCH AIM: To determine the feasibility of a metformin- versus-placebo definitive randomized clinical trial in women with low milk production and signs of insulin resistance. METHODS: Pilot trial criteria included: Mother 1-8 weeks postpartum (ideally 1-2 weeks), low milk production, and ≥1 insulin resistance sign; and singleton, healthy, term infant. Eligible mothers were randomly assigned 2:1 (metformin:placebo) and instructed in frequent milk removal for 28 days with option to stop at 14 days. RESULTS: From 02/2015 through 06/2016, we screened 114 women, completed baseline assessments on 46, and trialed 15 (median, 36 days postpartum). Comparing metformin-assigned ( n = 10) to placebo ( n = 5), 70% versus 80% continued to day 28; peak median change in milk output was +8 versus -58 mL/24 hr ( p = .31) and 80% peaked at Day 14 for both groups; 0% versus 20% desired to continue assigned drug after study completion; 44% versus 0% reported nausea/vomiting. Post-hoc, median peak change in milk output was +22 (metformin completers, n = 8) versus -58 mL/24 hr (placebo + non-completers, n = 7, p = .07). At baseline assessment, median milk production was significantly lower in those with ( n = 31), versus those without ( n = 15) signs of insulin resistance ( p = .002). CONCLUSIONS: Although results trend toward hypothesized direction, trial feasibility concerns include late enrollment and only 20% of metformin-assigned participants sustaining improved milk output to Day 28, with none perceiving metformin worthwhile. Better tools are needed to identify and treat metabolically-driven low milk production. Registered at ClinicalTrials.gov (NCT02179788) on 02/JUL/2014.

Extending TCGA queries to automatically identify analogous genomic data from dbGaP.

Data sharing is critical to advance genomic research by reducing the demand to collect new data by reusing and combining existing data and by promoting reproducible research. The Cancer Genome Atlas (TCGA) is a popular resource for individual-level genotype-phenotype cancer related data. The Database of Genotypes and Phenotypes (dbGaP) contains many datasets similar to those in TCGA. We have created a software pipeline that will allow researchers to discover relevant genomic data from dbGaP, based on matching TCGA metadata. The resulting research provides an easy to use tool to connect these two data sources.

The Relation between Breast Milk Sodium to Potassium Ratio and Maternal Report of a Milk Supply Concern.

We report that, among exclusively breastfeeding mothers at day 7 postpartum, those with milk supply concerns were significantly more likely to exhibit biochemical evidence of less progress toward mature lactation (elevated ratio of breast milk sodium to potassium concentration). Furthermore, an elevated ratio of breast milk sodium to potassium concentration was predictive of early weaning.

Protective coding variants in CFH and PELI3 and a variant near CTRB1 are associated with age-related macular degeneration†.

Although numerous common age-related macular degeneration (AMD) alleles have been discovered using genome-wide association studies, substantial disease heritability remains unexplained. We sought to identify additional common and rare variants associated with advanced AMD. A total of 4,332 cases and 25,268 controls of European ancestry from three different populations were genotyped using the Illumina Infinium HumanExome BeadChip. We performed meta-analyses to identify associations with common variants, and single variant and gene-based burden tests to identify rare variants. Two protective, low-frequency, non-synonymous variants were significantly associated with a decrease in AMD risk: A307V in PELI3 (odds ratio [OR] = 0.14, P = 4.3 × 10-10) and N1050Y in CFH (OR = 0.76, P = 6.2 × 10-12). The new variants have a large effect size, similar to some rare mutations we reported previously in a targeted sequencing study, which remain significant in this analysis: CFH R1210C (OR = 18.82, P = 3.5 × 10-07), C3 K155Q (OR = 3.27, P = 1.5 × 10-10) and C9 P167S (OR = 2.04, P = 2.8 × 10-07). We also identified a strong protective signal for a common variant (rs8056814) near CTRB1 associated with a decrease in AMD risk (logistic regression: OR = 0.71, P = 1.8 × 10-07). Suggestive protective loci were identified in the COL4A3 and APOH genes. Our results support the involvement of common and low-frequency protective variants in this vision-threatening condition. This study expands the roles of the innate immune pathway as well as the extracellular matrix and high-density lipoprotein pathways in the aetiology of AMD.

Long intergenic non-coding RNA expression signature in human breast cancer.

Breast cancer is a complex disease, characterized by gene deregulation. There is less systematic investigation of the capacity of long intergenic non-coding RNAs (lincRNAs) as biomarkers associated with breast cancer pathogenesis or several clinicopathological variables including receptor status and patient survival. We designed a two-stage study, including 1,000 breast tumor RNA-seq data from The Cancer Genome Atlas (TCGA) as the discovery stage, and RNA-seq data of matched tumor and adjacent normal tissue from 50 breast cancer patients as well as 23 normal breast tissue from healthy women as the replication stage. We identified 83 lincRNAs showing the significant expression changes in breast tumors with a false discovery rate (FDR) < 1% in the discovery dataset. Thirty-seven out of the 83 were validated in the replication dataset. Integrative genomic analyses suggested that the aberrant expression of these 37 lincRNAs was probably related with the expression alteration of several transcription factors (TFs). We observed a differential co-expression pattern between lincRNAs and their neighboring genes. We found that the expression levels of one lincRNA (RP5-1198O20 with Ensembl ID ENSG00000230615) were associated with breast cancer survival with P < 0.05. Our study identifies a set of aberrantly expressed lincRNAs in breast cancer.

Tissue-specific Co-expression of Long Non-coding and Coding RNAs Associated with Breast Cancer.

Inference of the biological roles of lncRNAs in breast cancer development remains a challenge. Here, we analyzed RNA-seq data in tumor and normal breast tissue samples from 18 breast cancer patients and 18 healthy controls and constructed a functional lncRNA-mRNA co-expression network. We revealed two distinctive co-expression patterns associated with breast cancer, reflecting different underlying regulatory mechanisms: (1) 516 pairs of lncRNA-mRNAs have differential co-expression pattern, in which the correlation between lncRNA and mRNA expression differs in tumor and normal breast tissue; (2) 291 pairs have dose-response co-expression pattern, in which the correlation is similar, but the expression level of lncRNA or mRNA differs in the two tissue types. We further validated our findings in TCGA dataset and annotated lncRNAs using TANRIC. One novel lncRNA, AC145110.1 on 8p12, was found differentially co-expressed with 127 mRNAs (including TOX4 and MAEL) in tumor and normal breast tissue and also highly correlated with breast cancer clinical outcomes. Functional enrichment and pathway analyses identified distinct biological functions for different patterns of co-expression regulations. Our data suggested that lncRNAs might be involved in breast tumorigenesis through the modulation of gene expression in multiple pathologic pathways.

Mapping rare, deleterious mutations in Factor H: Association with early onset, drusen burden, and lower antigenic levels in familial AMD.

The genetic architecture of age-related macular degeneration (AMD) involves numerous genetic variants, both common and rare, in the coding region of complement factor H (CFH). While these variants explain high disease burden in some families, they fail to explain the pathology in all. We selected families whose AMD was unexplained by known variants and performed whole exome sequencing to probe for other rare, highly penetrant variants. We identified four rare loss-of-function variants in CFH associated with AMD. Missense variant CFH 1:196646753 (C192F) segregated perfectly within a family characterized by advanced AMD and drusen temporal to the macula. Two families, each comprising a pair of affected siblings with extensive extramacular drusen, carried essential splice site variant CFH 1:196648924 (IVS6+1G>A) or missense variant rs139360826 (R175P). In a fourth family, missense variant rs121913058 (R127H) was associated with AMD. Most carriers had early onset bilateral advanced AMD and extramacular drusen. Carriers tended to have low serum Factor H levels, especially carriers of the splice variant. One missense variant (R127H) has been previously shown not to be secreted. The two other missense variants were produced recombinantly: compared to wild type, one (R175P) had no functional activity and the other (C192F) had decreased secretion.

Rare Variants in the Functional Domains of Complement Factor H Are Associated With Age-Related Macular Degeneration.

PURPOSE: Age-related macular degeneration (AMD) has a substantial genetic risk component, as evidenced by the risk from common genetic variants uncovered in the first genome-wide association studies. More recently, it has become apparent that rare genetic variants also play an independent role in AMD risk. We sought to determine if rare variants in complement factor H (CFH) played a role in AMD risk. METHODS: We had previously collected DNA from a large population of patients with advanced age-related macular degeneration (A-AMD) and controls for targeted deep sequencing of candidate AMD risk genes. In this analysis, we tested for an increased burden of rare variants in CFH in 1665 cases and 752 controls from this cohort. RESULTS: We identified 65 missense, nonsense, or splice-site mutations with a minor allele frequency ≤ 1%. Rare variants with minor allele frequency ≤ 1% (odds ratio [OR] = 1.5, P = 4.4 × 10⁻²), 0.5% (OR = 1.6, P = 2.6 × 10⁻²), and all singletons (OR = 2.3, P = 3.3 × 10⁻²) were enriched in A-AMD cases. Moreover, we observed loss-of-function rare variants (nonsense, splice-site, and loss of a conserved cysteine) in 10 cases and serum levels of FH were decreased in all 5 with an available sample (haploinsufficiency). Further, rare variants in the major functional domains of CFH were increased in cases (OR = 3.2; P = 1.4 × 10⁻³) and the magnitude of the effect correlated with the disruptive nature of the variant, location in an active site, and inversely with minor allele frequency. CONCLUSIONS: In this large A-AMD cohort, rare variants in the CFH gene were enriched and tended to be located in functional sites or led to low serum levels. These data, combined with those indicating a similar, but even more striking, increase in rare variants found in CFI, strongly implicate complement activation in A-AMD etiopathogenesis as CFH and CFI interact to inhibit the alternative pathway.

Systems biology evaluation of cell-free amniotic fluid transcriptome of term and preterm infants to detect fetal maturity.

BACKGROUND: Amniotic fluid (AF) is a proximal fluid to the fetus containing higher amounts of cell-free fetal RNA/DNA than maternal serum, thereby making it a promising source for identifying novel biomarkers that predict fetal development and organ maturation. Our aim was to compare AF transcriptomic profiles at different time points in pregnancy to demonstrate unique genetic signatures that would serve as potential biomarkers indicative of fetal maturation. METHODS: We isolated AF RNA from 16 women at different time points in pregnancy: 4 from 18 to 24 weeks, 6 from 34 to 36 weeks, and 6 from 39 to 40 weeks. RNA-sequencing was performed on cell-free RNA. Gene expression and splicing analyses were performed in conjunction with cell-type and pathway predictions. RESULTS: Sample-level analysis at different time points in pregnancy demonstrated a strong correlation with cell types found in the intrauterine environment and fetal respiratory, digestive and external barrier tissues of the fetus, using high-confidence cellular molecular markers. While some RNAs and splice variants were present throughout pregnancy, many transcripts were uniquely expressed at different time points in pregnancy and associated with distinct neonatal co-morbidities (respiratory distress and gavage feeding), indicating fetal immaturity. CONCLUSION: The AF transcriptome exhibits unique cell/organ-selective expression patterns at different time points in pregnancy that can potentially identify fetal organ maturity and predict neonatal morbidity. Developing novel biomarkers indicative of the maturation of multiple organ systems can improve upon our current methods of fetal maturity testing which focus solely on the lung, and will better inform obstetrical decisions regarding delivery timing.