Inference of the human polyadenylation code.

Motivation: Processing of transcripts at the 3'-end involves cleavage at a polyadenylation site followed by the addition of a poly(A)-tail. By selecting which site is cleaved, the process of alternative polyadenylation enables genes to produce transcript isoforms with different 3'-ends. To facilitate the identification and treatment of disease-causing mutations that affect polyadenylation and to understand the sequence determinants underlying this regulatory process, a computational model that can accurately predict polyadenylation patterns from genomic features is desirable. Results: Previous works have focused on identifying candidate polyadenylation sites and classifying tissue-specific sites. By training on how multiple sites in genes are competitively selected for polyadenylation from 3'-end sequencing data, we developed a deep learning model that can predict the tissue-specific strength of a polyadenylation site in the 3' untranslated region of the human genome given only its genomic sequence. We demonstrate the model's broad utility on multiple tasks, without any application-specific training. The model can be used to predict which polyadenylation site is more likely to be selected in genes with multiple sites. It can be used to scan the 3' untranslated region to find candidate polyadenylation sites. It can be used to classify the pathogenicity of variants near annotated polyadenylation sites in ClinVar. It can also be used to anticipate the effect of antisense oligonucleotide experiments to redirect polyadenylation. We provide analysis on how different features affect the model's predictive performance and a method to identify sensitive regions of the genome at the single-based resolution that can affect polyadenylation regulation. Supplementary information: Supplementary data are available at Bioinformatics online.

Deep learning of the tissue-regulated splicing code.

MOTIVATION: Alternative splicing (AS) is a regulated process that directs the generation of different transcripts from single genes. A computational model that can accurately predict splicing patterns based on genomic features and cellular context is highly desirable, both in understanding this widespread phenomenon, and in exploring the effects of genetic variations on AS. METHODS: Using a deep neural network, we developed a model inferred from mouse RNA-Seq data that can predict splicing patterns in individual tissues and differences in splicing patterns across tissues. Our architecture uses hidden variables that jointly represent features in genomic sequences and tissue types when making predictions. A graphics processing unit was used to greatly reduce the training time of our models with millions of parameters. RESULTS: We show that the deep architecture surpasses the performance of the previous Bayesian method for predicting AS patterns. With the proper optimization procedure and selection of hyperparameters, we demonstrate that deep architectures can be beneficial, even with a moderately sparse dataset. An analysis of what the model has learned in terms of the genomic features is presented.

In vivo optical imaging of tumor and microvascular response to ionizing radiation.

Radiotherapy is a widely used cancer treatment. However, understanding how ionizing radiation affects tumor cells and their vasculature, particularly at cellular, subcellular, genetic, and protein levels, has been limited by an inability to visualize the response of these interdependent components within solid tumors over time and in vivo. Here we describe a new preclinical experimental platform combining intravital multimodal optical microscopy for cellular-level longitudinal imaging, a small animal x-ray microirradiator for reproducible spatially-localized millimeter-scale irradiations, and laser-capture microdissection of ex vivo tissues for transcriptomic profiling. Using this platform, we have developed new methods that exploit the power of optically-enabled microscopic imaging techniques to reveal the important role of the tumor microvasculature in radiation response of tumors. Furthermore, we demonstrate the potential of this preclinical platform to study quantitatively--with cellular and sub-cellular details--the spatio-temporal dynamics of the biological response of solid tumors to ionizing radiation in vivo.

Monte Carlo simulation on a gold nanoparticle irradiated by electron beams.

This study investigated the secondary electron production from a gold nanoparticle (GNP) irradiated by monoenergetic electron beams using Monte Carlo (MC) simulation. Spherical GNPs with diameters of 2, 50 and 100 nm in water were irradiated by monoenergetic electron beams with energies equal to 50 keV, 250 keV, 1 MeV and 4 MeV. MC simulations were performed using the Geant4 toolkit to determine the energy of the secondary electrons emitted from the GNPs. The mean effective range and deflection angle of the secondary electrons were tracked. Energy depositions inside and outside the nanoparticles due to the secondary electrons were also calculated. For comparisons, simulations were repeated by replacing the GNPs with water. Our results show that the mean effective range of secondary electrons increased with an increase of the GNP size and electron beam energy. For the electron beam energy and GNP size used in this study, the mean effective range was 0.5-15 µm outside the nanoparticle, which is approximately within the dimension of a living cell. The mean deflection angles varied from 78 to 83 degrees as per our MC results. The proportion of energy deposition inside the GNP versus that outside increased with the GNP size. This is different from the results obtained from a previous study using photon beams. The secondary electron energy deposition ratio (energy deposition for GNP/energy deposition for water) was found to be highest for the smallest GNP of 2 nm diameter in this study. For the energy deposited by the secondary electron, we concluded that the addition of GNPs can increase the secondary electron energy deposition in water, though most of the energy was self-absorbed by the large nanoparticles (50 and 100 nm). In addition, an electron source in the presence of GNPs does not seem to be better than photons as the yield of secondary electrons per unit mass of gold is less than water.

Irradiation of gold nanoparticles by x-rays: Monte Carlo simulation of dose enhancements and the spatial properties of the secondary electrons production.

PURPOSE: The aim of this study is to understand the characteristics of secondary electrons generated from the interaction of gold nanoparticles (GNPs) with x-rays as a function of nanoparticle size and beam energy and thereby further the understanding of GNP-enhanced radiotherapy. METHODS: The effective range, deflection angle, dose deposition, energy, and interaction processes of electrons produced from the interaction of x-rays with a GNP were calculated by Monte Carlo simulations. The GEANT4 code was used to simulate and track electrons generated from a 2, 50, and 100 nm diameter GNP when it is irradiated with a 50 kVp, 250 kVp, cobalt-60, and 6 MV photon beam in water. RESULTS: When a GNP was present, depending on the beam energies used, secondary electron production was increased by 10- to 2000-fold compared to an absence of a GNP. Low-energy photon beams were much more efficient at interacting with the GNP by two to three orders of magnitude compared to MV energies and increased the deflection angle. GNPs with larger diameters also contributed more dose. The majority of the energy deposition was outside the GNP, rather than self-absorbed by the nanoparticle. The mean effective range of electron tracks for the beams tested ranged from approximately 3 microm to 1 mm. CONCLUSIONS: These simulated results yield important insights concerning the spatial distributions and elevated dose in GNP-enhanced radiotherapy. The authors conclude that the irradiation of GNP at lower photon energies will be more efficient for cell killing. This conclusion is consistent with published studies.

Dosimetric variation due to the photon beam energy in the small-animal irradiation: a Monte Carlo study.

PURPOSE: The impact of photon beam energy and tissue heterogeneities on dose distributions and dosimetric characteristics such as point dose, mean dose, and maximum dose was investigated in the context of small-animal irradiation using Monte Carlo simulations based on the EGSnrc code. METHODS: Three Monte Carlo mouse phantoms, namely, heterogeneous, homogeneous, and bone homogeneous were generated based on the same mouse computed tomography image set. These phantoms were generated by overriding the tissue type of none of the voxels (heterogeneous), all voxels (homogeneous), and only the bone voxels (bone homogeneous) to that of soft tissue. Phase space files of the 100 and 225 kVp photon beams based on a small-animal irradiator (XRad225Cx, Precision X-Ray Inc., North Branford, CT) were generated using BEAMnrc. A 360 degrees photon arc was simulated and three-dimensional (3D) dose calculations were carried out using the DOSXYZnrc code through DOSCTP in the above three phantoms. For comparison, the 3D dose distributions, dose profiles, mean, maximum, and point doses at different locations such as the isocenter, lung, rib, and spine were determined in the three phantoms. RESULTS: The dose gradient resulting from the 225 kVp arc was found to be steeper than for the 100 kVp arc. The mean dose was found to be 1.29 and 1.14 times higher for the heterogeneous phantom when compared to the mean dose in the homogeneous phantom using the 100 and 225 kVp photon arcs, respectively. The bone doses (rib and spine) in the heterogeneous mouse phantom were about five (100 kVp) and three (225 kVp) times higher when compared to the homogeneous phantom. However, the lung dose did not vary significantly between the heterogeneous, homogeneous, and bone homogeneous phantom for the 225 kVp compared to the 100 kVp photon beams. CONCLUSIONS: A significant bone dose enhancement was found when the 100 and 225 kVp photon beams were used in small-animal irradiation. This dosimetric effect, due to the presence of the bone heterogeneity, was more significant than that due to the lung heterogeneity. Hence, for kV photon energies of the range used in small-animal irradiation, the increase of the mean and bone dose due to the photoelectric effect could be a dosimetric concern.

Optimized speckle variance OCT imaging of microvasculature.

We optimize speckle variance optical coherence tomography (svOCT) imaging of microvasculature in high and low bulk tissue motion scenarios. To achieve a significant level of image contrast, frame rates must be optimized such that tissue displacement between frames is less than the beam radius. We demonstrate that higher accuracy estimates of speckle variance can enhance the detection of capillaries. These findings are illustrated in vivo by imaging the dorsal window chamber model (low bulk motion). We also show svOCT imaging of the nonstabilized finger (high bulk motion), using optimized imaging parameters, demonstrating better vessel detection than Doppler OCT.

Doppler optical coherence tomography for interventional cardiovascular guidance: in vivo feasibility and forward-viewing probe flow phantom demonstration.

We demonstrate the potential of a forward-looking Doppler optical coherence tomography (OCT) probe for color flow imaging in several commonly seen narrowed artery morphologies. As a proof of concept, we present imaging results of a surgically exposed thrombotic occlusion model that was imaged superficially to demonstrate that Doppler OCT can identify flow within the recanalization channels of a blocked artery. We present Doppler OCT images in which the flow is nearly antiparallel to the imaging direction. These images are acquired using a flexible 2.2-mm-diam catheter that used electrostatic actuation to scan up to 30 deg ahead of the distal end. Doppler OCT images of physiologically relevant flow phantoms consisting of small channels and tapered entrance geometries are demonstrated.

In vivo endoscopic multi-beam optical coherence tomography.

A multichannel optical coherence tomography (multi-beam OCT) system and an in vivo endoscopic imaging probe were developed using a swept-source OCT system. The distal optics were micro-machined to produce a high numerical aperture, multi-focus fibre optic array. This combination resulted in a transverse design resolution of <10 microm full width half maximum (FWHM) throughout the entire imaging range, while also increasing the signal intensity within the focus of the individual channels. The system was used in a pre-clinical rabbit study to acquire in vivo structural images of the colon and ex vivo images of the oesophagus and trachea. A good correlation between the structural multi-beam OCT images and H&E histology was achieved, demonstrating the feasibility of this high-resolution system and its potential for in vivo human endoscopic imaging.

Dosimetry of oblique tangential photon beams calculated by superposition/convolution algorithms: a Monte Carlo evaluation.

Although there are many works on evaluating dose calculations of the anisotropic analytical algorithm (AAA) using various homogeneous and heterogeneous phantoms, related work concerning dosimetry due to tangential photon beam is lacking. In this study, dosimetry predicted by the AAA and collapsed cone convolution (CCC) algorithm was evaluated using the tangential photon beam and phantom geometry. The photon beams of 6 and 15 MV with field sizes of 4 × 4 (or 7 × 7), 10 × 10 and 20 × 20 cm², produced by a Varian 21 EX linear accelerator, were used to test performances of the AAA and CCC using Monte Carlo (MC) simulation (EGSnrc-based code) as a benchmark. Horizontal dose profiles at different depths, phantom skin profiles (i.e., vertical dose profiles at a distance of 2 mm from the phantom lateral surface), gamma dose distributions, and dose-volume histograms (DVHs) of skin slab were determined. For dose profiles at different depths, the CCC agreed better with doses in the air-phantom region, while both the AAA and CCC agreed well with doses in the penumbra region, when compared to the MC. Gamma evaluations between the AAA/CCC and MC showed that deviations of 2D dose distribution occurred in both beam edges in the phantom and air-phantom interface. Moreover, the gamma dose deviation is less significant in the air-phantom interface than the penumbra. DVHs of skin slab showed that both the AAA and CCC underestimated the width of the dose drop-off region for both the 6 and 15 MV photon beams. When the gantry angle was 0°, it was found that both the AAA and CCC overestimated doses in the phantom skin profiles compared to the MC, with various photon beam energies and field sizes. The mean dose differences with doses normalized to the prescription point for the AAA and CCC were respectively: 7.6% ± 2.6% and 2.1% ± 1.3% for a 10 × 10 cm2 field, 6 MV; 16.3%± 2.1% and 6.7% ± 2.1% for a 20 × 20 cm2 field, 6 MV; 5.5% ± 1.2% and 1.7% ± 1.4% for a 10 × 10 cm2, 15 MV; 18.0% ± 1.3% and 8.3% ± 1.8% for a 20 × 20 cm², 15 MV. However, underestimations of doses in the phantom skin profile were found with small fields of 4 × 4 and 7 × 7 cm² for the 6 and 15 MV photon beams, respectively, when the gantry was turned 5° anticlockwise. As surface dose with tangential photon beam geometry is important in some radiation treatment sites such as breast, chest wall and sarcoma, it is found that neither of the treatment planning system algorithms can predict the dose well at depths shallower than 2 mm. The dosimetry data and beam and phantom geometry in this study provide a better knowledge of a dose calculation algorithm in tangential-like irradiation.

High-power wavelength-swept laser in Littman telescope-less polygon filter and dual-amplifier configuration for multichannel optical coherence tomography.

We report a high-power wavelength-swept laser source for multichannel optical coherence tomography (OCT) imaging. Wavelength tuning is performed by a compact telescope-less polygon-based filter in Littman arrangement. High output power is achieved by incorporating two serial semiconductor optical amplifiers in the laser cavity in Fourier domain mode-locked configuration. The measured wavelength tuning range of the laser is 111 nm centered at 1329 nm, coherence length of 5.5 mm, and total average output power of 131 mW at 43 kHz sweeping rate. Multichannel simultaneous OCT imaging at an equivalent A-scan rate of 258 kHz is demonstrated.

Variations of lung density and geometry on inhomogeneity correction algorithms: a Monte Carlo dosimetric evaluation.

This work contributed the following new information to the study of inhomogeneity correction algorithm: (1) Evaluation of lung dose calculation methods as a function of lung relative electron density (rhoe,lung) and treatment geometry and (2) comparison of doses calculated using the collapsed cone convolution (CCC) and adaptive convolution (AC) in lung using the Monte Carlo (MC) simulation with the EGSnrc-based code. The variations of rhoe,lung and geometry such as the position and dimension of the lung were studied with different photon beam energies and field sizes. Three groups of inhomogeneous lung phantoms, namely, "slab," "column," and "cube," with different positions, volumes, and shapes of lung in water as well as clinical computed tomography lung images were used. The rhoe,lung in each group of phantoms vary from 0.05 to 0.7. 6 and 18 MV photon beams with small (4 x 4 cm2) and medium (10 x 10 cm2) field sizes produced by a Varian 21 EX linear accelerator were used. This study reveals that doses in the inhomogeneous lung calculated by the CCC match well with those by AC within +/- 1%, indicating that the AC, with an advantage of shorter computing times (three to four times shorter than CCC), is a good substitute for CCC. Comparing the CCC and AC to MC in general, significant dose deviations are found when the rhoe,lung is < or =0.3. The degree of deviation depends on the photon beam energy and field size and is relatively large when high-energy photon beams with small fields are used. For penumbra widths (20%-80%), the CCC and AC agree well with MC for the slab and cube phantoms with the lung volumes at the central beam axis (CAX). However, deviations (>2 mm) occur in the column phantoms, with two lung volumes separated by a unit density column along the CAX in the middle using the 18 MV beam with 4 x 4 cm2 field for rhoe,lung < or =0.1. This study provides new dosimetric data to evaluate the impact of the variations of rhoe,lung and geometry on dose calculations in inhomogeneous media using CCC and AC.

Monte Carlo simulation of MOSFET dosimeter for electron backscatter using the GEANT4 code.

The aim of this study is to investigate the influence of the body of the metal-oxide-semiconductor field effect transistor (MOSFET) dosimeter in measuring the electron backscatter from lead. The electron backscatter factor (EBF), which is defined as the ratio of dose at the tissue-lead interface to the dose at the same point without the presence of backscatter, was calculated by the Monte Carlo simulation using the GEANT4 code. Electron beams with energies of 4, 6, 9, and 12 MeV were used in the simulation. It was found that in the presence of the MOSFET body, the EBFs were underestimated by about 2%-0.9% for electron beam energies of 4-12 MeV, respectively. The trend of the decrease of EBF with an increase of electron energy can be explained by the small MOSFET dosimeter, mainly made of epoxy and silicon, not only attenuated the electron fluence of the electron beam from upstream, but also the electron backscatter generated by the lead underneath the dosimeter. However, this variation of the EBF underestimation is within the same order of the statistical uncertainties as the Monte Carlo simulations, which ranged from 1.3% to 0.8% for the electron energies of 4-12 MeV, due to the small dosimetric volume. Such small EBF deviation is therefore insignificant when the uncertainty of the Monte Carlo simulation is taken into account. Corresponding measurements were carried out and uncertainties compared to Monte Carlo results were within +/- 2%. Spectra of energy deposited by the backscattered electrons in dosimetric volumes with and without the lead and MOSFET were determined by Monte Carlo simulations. It was found that in both cases, when the MOSFET body is either present or absent in the simulation, deviations of electron energy spectra with and without the lead decrease with an increase of the electron beam energy. Moreover, the softer spectrum of the backscattered electron when lead is present can result in a reduction of the MOSFET response due to stronger recombination in the SiO2 gate. It is concluded that the MOSFET dosimeter performed well for measuring the electron backscatter from lead using electron beams. The uncertainty of EBF determined by comparing the results of Monte Carlo simulations and measurements is well within the accuracy of the MOSFET dosimeter (< +/- 4.2%) provided by the manufacturer.

Evaluation of the effect of patient dose from cone beam computed tomography on prostate IMRT using Monte Carlo simulation.

The aim of this study is to evaluate the impact of the patient dose due to the kilovoltage cone beam computed tomography (kV-CBCT) in a prostate intensity-modulated radiation therapy (IMRT). The dose distributions for the five prostate IMRTs were calculated using the Pinnacle treatment planning system. To calculate the patient dose from CBCT, phase-space beams of a CBCT head based on the ELEKTA x-ray volume imaging system were generated using the Monte Carlo BEAMnr code for 100, 120, 130, and 140 kVp energies. An in-house graphical user interface called DOSCTP (DOSXYZnrc-based) developed using MATLAB was used to calculate the dose distributions due to a 360 degrees photon arc from the CBCT beam with the same patient CT image sets as used in Pinnacle. The two calculated dose distributions were added together by setting the CBCT doses equal to 1%, 1.5%, 2%, and 2.5% of the prescription dose of the prostate IMRT. The prostate plan and the summed dose distributions were then processed in the CERR platform to determine the dose-volume histograms (DVHs) of the regions of interest. Moreover, dose profiles along the x- and y-axes crossing the isocenter with and without addition of the CBCT dose were determined. It was found that the added doses due to CBCT are most significant at the femur heads. Higher doses were found at the bones for a relatively low energy CBCT beam such as 100 kVp. Apart from the bones, the CBCT dose was observed to be most concentrated on the anterior and posterior side of the patient anatomy. Analysis of the DVHs for the prostate and other critical tissues showed that they vary only slightly with the added CBCT dose at different beam energies. On the other hand, the changes of the DVHs for the femur heads due to the CBCT dose and beam energy were more significant than those of rectal and bladder wall. By analyzing the vertical and horizontal dose profiles crossing the femur heads and isocenter, with and without the CBCT dose equal to 2% of the prescribed dose, it was found that there is about a 5% increase of dose at the femur head. Still, such an increase in the femur head dose is well below the dose limit of the bone in our IMRT plans. Therefore, under these dose fractionation conditions, it is concluded that, though CBCT causes a higher dose deposited at the bones, there may be no significant effect in the DVHs of critical tissues in the prostate IMRT.

Treatment planning for a small animal using Monte Carlo simulation.

The development of a small animal model for radiotherapy research requires a complete setup of customized imaging equipment, irradiators, and planning software that matches the sizes of the subjects. The purpose of this study is to develop and demonstrate the use of a flexible in-house research environment for treatment planning on small animals. The software package, called DOSCTP, provides a user-friendly platform for DICOM computed tomography-based Monte Carlo dose calculation using the EGSnrcMP-based DOSXYZnrc code. Validation of the treatment planning was performed by comparing the dose distributions for simple photon beam geometries calculated through the Pinnacle3 treatment planning system and measurements. A treatment plan for a mouse based on a CT image set by a 360-deg photon arc is demonstrated. It is shown that it is possible to create 3D conformal treatment plans for small animals with consideration of inhomogeneities using small photon beam field sizes in the diameter range of 0.5-5 cm, with conformal dose covering the target volume while sparing the surrounding critical tissue. It is also found that Monte Carlo simulation is suitable to carry out treatment planning dose calculation for small animal anatomy with voxel size about one order of magnitude smaller than that of the human.