EpICC: A Bayesian neural network model with uncertainty correction for a more accurate classification of cancer.

Accurate classification of cancers into their types and subtypes holds the key for choosing the right treatment strategy and can greatly impact patient well-being. However, existence of large-scale variations in the molecular processes driving even a single type of cancer can make accurate classification a challenging problem. Therefore, improved and robust methods for classification are absolutely critical. Although deep learning-based methods for cancer classification have been proposed earlier, they all provide point estimates for predictions without any measure of confidence and thus, can fall short in real-world applications where key decisions are to be made based on the predictions of the classifier. Here we report a Bayesian neural network-based model for classification of cancer types as well as sub-types from transcriptomic data. This model reported a measure of confidence with each prediction through analysis of epistemic uncertainty. We incorporated an uncertainty correction step with the Bayesian network-based model to greatly enhance prediction accuracy of cancer types (> 97% accuracy) and sub-types (> 80%). Our work suggests that reporting uncertainty measure with each classification can enable more accurate and informed decision-making that can be highly valuable in clinical settings.

Adsorption of ammonia on graphene.

We report on experimental studies of NH3 adsorption/desorption on graphene surfaces. The study employs bottom-gated graphene field effect transistors supported on Si/SiO2 substrates. Detection of NH3 occurs through the shift of the source-drain resistance maximum ('Dirac peak') with the gate voltage. The observed shift of the Dirac peak toward negative gate voltages in response to NH3 exposure is consistent with a small charge transfer (f approximately 0.068 +/- 0.004 electrons per molecule at pristine sites) from NH3 to graphene. The desorption kinetics involves a very rapid loss of NH3 from the top surface and a much slower removal from the bottom surface at the interface with the SiO2 that we identify with a Fickian diffusion process.

n-Type behavior of graphene supported on Si/SiO(2) substrates.

Results are presented from an experimental and theoretical study of the electronic properties of back-gated graphene field effect transistors (FETs) on Si/SiO(2) substrates. The excess charge on the graphene was observed by sweeping the gate voltage to determine the charge neutrality point in the graphene. Devices exposed to laboratory environment for several days were always found to be initially p-type. After approximately 20 h at 200 degrees C in approximately 5 x 10(-7) Torr vacuum, the FET slowly evolved to n-type behavior with a final excess electron density on the graphene of approximately 4 x 10(12) e/cm(2). This value is in excellent agreement with our theoretical calculations on SiO(2), where we have used molecular dynamics to build the SiO(2) structure and then density functional theory to compute the electronic structure. The essential theoretical result is that the SiO(2) has a significant surface state density just below the conduction band edge that donates electrons to the graphene to balance the chemical potential at the interface. An electrostatic model for the FET is also presented that produces an expression for the gate bias dependence of the carrier density.