Predicting Survival Outcomes in the Presence of Unlabeled Data

Many clinical studies require the follow-up of patients over time. This is challenging: apart from frequently observed drop-out, there are often also organizational and financial challenges, which can lead to reduced data collection and, in turn, can complicate subsequent analyses. In contrast, there is often plenty of baseline data available of patients with similar characteristics and background information, e.g., from patients that fall outside the study time window. In this article, we investigate whether we can benefit from the inclusion of such unlabeled data instances to predict accurate survival times. In other words, we introduce a third level of supervision in the context of survival analysis, apart from fully observed and censored instances, we also include unlabeled instances. We propose three approaches to deal with this novel setting and provide an empirical comparison over fifteen real-life clinical and gene expression survival datasets. Our results demonstrate that all approaches are able to increase the predictive performance over independent test data. We also show that integrating the partial supervision provided by censored data in a semi-supervised wrapper approach generally provides the best results, often achieving high improvements, compared to not using unlabeled data.

     

Low colostrum yield in Jersey cattle and potential risk factors

Consumption of an adequate volume of high-quality colostrum is vital to a dairy calf's ability to survive and become a productive herd member. However, some dairy herds have reported a deficiency of colostrum production, which ranges from a low volume to no colostrum produced, by cows during fall and winter. Little information regarding this phenomenon exists. The purpose of this study was to characterize the syndrome and identify potential risk factors for low colostrum yield. A 2,500-cow Jersey dairy farm was enrolled in a prospective cohort study in May 2016, to evaluate possible effects of photoperiod, temperature, and cow factors on colostrum production. Dairy personnel were trained to collect, weigh, and evaluate colostrum quality. Information on parity, previous lactation length, previous 305-d mature-equivalent milk production, and dry period length were collected through the farm's dairy management software. Weather and photoperiod data were also collected. Over the year of enrollment, 2,988 eligible cows calved and had colostrum weights recorded and 38% were primiparous (n = 1,143), 25% were in their second lactation (n = 752), and 37% were in their third or greater lactation (n = 1,093). The overall average colostrum yield was 6.6 kg/cow in June 2016, 2.5 kg/cow in December 2016, and 4.8 kg/cow in May 2017. Multiparous cows had a larger decline in colostrum production between June and December (6.6 to 1.3 kg/cow) compared with primiparous animals (6.5 to 4.2 kg/cow). Overall, average colostrum production decreased by 0.17 kg/cow per week during this time, 0.22 kg for multiparous cows and 0.08 kg for primiparous cows. A logistic regression model was constructed for all cows to evaluate effects of cow factors on low colostrum production (<2.7 kg at first milking). Dry period length, calf sex, singleton or twin, age at freshening, month of calving and previous lactation length were significantly associated with the probability of low colostrum yield (<2.7 kg at first milking). A cross-correlation function analysis between the time series for colostrum yield and photoperiod revealed a high correlation at the time of calving and 1 mo prior, particularly for multiparous cows. A pedigree analysis showed that extreme colostrum yield (low vs. high) followed some sire lines. Low colostrum production in this herd could have an economic effect on the dairy and calf health and appears to have a strong seasonal and, potentially, a genetic component.     

Nonlinear dynamic response of tall buildings considering structure–soil–structure effects

Numerous studies have shown that the interaction between adjacent buildings can result in changes in nonlinear dynamic response of structures, damage, and performance level, depending on the dynamic specifications of structures involved and the frequency content of the input motion. To study these effects, finite element method is used for the analytical investigations, and total soil–foundation–structure system is modeled all together. For modeling purposes and in order to realize the effects of the adjacent buildings on the dynamic response, two buildings, namely, 15‐story and 30‐story tall buildings, which were separated by distances of 1/4 and 1/8 of the width of the foundation and were located on hard and soft soil profiles, were considered. It was concluded that in the case where the soil and structure's periods were near to each other, the interaction of adjacent structures on increasing nonlinear responses (displacement and interstory drift) and structural damage indexes was noticeable and therefore was not negligible. Whereas in the case where periods are distant from each other, the interaction of adjacent buildings has a decreasing effect on damage indexes and nonlinear responses and therefore was negligible. Copyright © 2011 John Wiley & Sons, Ltd.     

Prognostic factors in bone marrow transplantation for beta thalassemia major: experiences from Iran

This study concerns the effects of several pre-transplant features on outcome for patients with beta thalassemia major who underwent bone marrow transplantation (BMT). Seventy patients with beta thalassemia major underwent bone marrow transplantation during the period 1991-1997 in Shariati Hospital in Tehran, Iran. The survival and rejection curves levelled off at 8 and 18 months after transplantation at 82.6% and 11.4%, respectively. Pre-transplant clinical features (age, serum ferritin, portal fibrosis, hepatomegaly and quality of chelation therapy) were examined for their effects on survival and recurrence of thalassemia in this group of patients who were less than 16 years old. Increasing age, presence of portal fibrosis and increasing serum ferritin were significantly associated with reduced probability of survival (P = 0.0047, P = 0.016 and P = 0.024, respectively). Hepatomegaly and inadequate pre-transplant chelation therapy which were documented as poor prognostic factors in previous studies, were not evaluable in this study. We also showed the benefits of transplanting more than 5.5 x 10(8)/kg cells in this group of patients with no increase in complications.     

Effects of dopants and copolymerization on Schottky barriers of polypyrrole and polyindole/metal interfaces

Schottky-barrier diode devices were fabricated in a sandwich configuration with poly(pyrrole-co-indole) copolymer semiconducting films prepared by electropolymerization. Effect of different dopants of ClO₄ ^?, BF₄ ^?, C₇H₇SO₃ ^? and [Fe(CN)₆]^3? on the electronic properties of the fabricated devices was followed using Ag, In, Al and Cu metal junctions. Current?Cvoltage and capacitance?Cvoltage characteristics were recorded for making a comparative evaluation of the electronic and junction properties of the devices. The electrical characteristics of the junctions were analyzed based on the standard thermionic emission theory. Polymer doped by ClO₄ ^? showed lower reverse saturation currents and ideality factor but higher potential barriers and rectification ratios. Effect of dopant ions and copolymerization on the optical band gaps (E _g) of the films were investigated and the optical transmissions of the doped copolymer films were measured in the wavelength range of 250?C900?nm. It was shown that the energy gap of copolymers laid between those of corresponding homopolymers and polyindole (PIN) doped by [Fe(CN₆)]^?3 had E _g less than that of polymer doped by other anions whereas E _g of polypyrrole was independent of dopant ions. Also, the morphology of the polymeric films revealed the surface of the PIN doped with ClO₄ ^? was very smooth which created a good contact with indium metal.     

Growth and Luminescence of Polytypic InP on Epitaxial Graphene

Van der Waals epitaxy is an attractive alternative to direct heteroepitaxy where the forced coherency at the interface cannot sustain large differences in lattice parameters and thermal expansion coefficients between the substrate and the epilayer. Herein, the growth of monocrystalline InP on Ge and SiO₂/Si substrates using graphene as an interfacial layer is demonstrated. Micrometer‐sized InP crystals are found to grow with interfaces of high crystalline quality and with different degrees of coalescence depending on the growth conditions. Some InP crystals exhibit a polytypic structure, consisting of alternating zinc‐blende and wurtzite phases, forming a type‐II homojunction with well (barrier) width of about 10 nm. The optical properties, investigated using room temperature nano‐cathodoluminescence, indicate the signatures of the direct optical transitions at 1.34 eV across the gap of the zinc‐blende phase and the indirect transitions at ≈ 1.31 eV originating from the alternating zinc‐blende and wurtzite phases. Additionally, the InP nanorods, found growing mainly on the graphene/SiO₂/Si substrate, show optical transition across the gap of the wurtzite phase at ≈ 1.42 eV. This demonstration of InP growth on graphene and the correlative study between the structure and optical properties pave the way to develop hybrid structures for potential applications in integrated photonic and optoelectronic devices.