Dataset Information

Machine learning methodology for high throughput personalized neutron dose reconstruction in mixed neutron?+?photon exposures.

ABSTRACT: We implemented machine learning in the radiation biodosimetry field to quantitatively reconstruct neutron doses in mixed neutron?+?photon exposures, which are expected in improvised nuclear device detonations. Such individualized reconstructions are crucial for triage and treatment because neutrons are more biologically damaging than photons. We used a high-throughput micronucleus assay with automated scanning/imaging on lymphocytes from human blood ex-vivo irradiated with 44 different combinations of 0-4 Gy neutrons and 0-15 Gy photons (542 blood samples), which include reanalysis of past experiments. We developed several metrics that describe micronuclei/cell probability distributions in binucleated cells, and used them as predictors in random forest (RF) and XGboost machine learning analyses to reconstruct the neutron dose in each sample. The probability of "overfitting" was minimized by training both algorithms with repeated cross-validation on a randomly-selected subset of the data, and measuring performance on the rest. RF achieved the best performance. Mean R² for actual vs. reconstructed neutron doses over 300 random training/testing splits was 0.869 (range 0.761 to 0.919) and root mean squared error was 0.239 (0.195 to 0.351) Gy. These results demonstrate the promising potential of machine learning to reconstruct the neutron dose component in clinically-relevant complex radiation exposure scenarios.

SUBMITTER: Shuryak I

PROVIDER: S-EPMC7889851 | biostudies-literature | 2021 Feb

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Machine learning methodology for high throughput personalized neutron dose reconstruction in mixed neutron + photon exposures.

Shuryak Igor I Turner Helen C HC Pujol-Canadell Monica M Perrier Jay R JR Garty Guy G Brenner David J DJ

Scientific reports 20210217 1

We implemented machine learning in the radiation biodosimetry field to quantitatively reconstruct neutron doses in mixed neutron + photon exposures, which are expected in improvised nuclear device detonations. Such individualized reconstructions are crucial for triage and treatment because neutrons are more biologically damaging than photons. We used a high-throughput micronucleus assay with automated scanning/imaging on lymphocytes from human blood ex-vivo irradiated with 44 different combinati ...[more]

PMID: 33597632

Similar Datasets

Project description:In the possible event of a detonation of an improvised nuclear device (IND), the immediate radiation would consist of both photons (gamma rays) and neutrons. Since neutrons generally have a high relative biological effectiveness (RBE) for most physiological end points, it is important to understand the effect that neutrons would have on the biodosimetry methods that are being developed for medical triage purposes. We previously compared the transcriptomic response in human blood after neutron and photon irradiation. In this study, we analyzed the effect of mixed-field-neutron-photon radiation on gene expression responses in human peripheral blood, to elucidate the neutron contribution in the setting of a radiation exposure from an IND detonation. We used four combinations of mixed neutron-photon exposures, with increasing percentages of neutrons, to a cumulative dose of 3 Gy. The mixed-field exposures consisted of 0%, 5%, 15% and 25% of neutrons, where 0% corresponds to 3 Gy of pure X rays. A maximum neutron exposure, corresponding to 83% neutrons (0.75 Gy) was also used in the study. Increases were observed in both the number and expression level of genes, with increasing percentages of neutrons from 0% to 25% in the mixed-field exposures. Gene ontology analysis showed an overall predominance of TP53 signaling among upregulated genes across all exposures. Some TP53 regulated genes, such as EDA2R, GDF15 and VWCE, demonstrated increased expression with increasing neutron percentages in mixed-field exposures. Immune response, specifically natural-killer-cell mediated signaling, was the most significant biological process associated with downregulated genes. We observed significant suppression of T-cell-mediated signaling in mixed-field exposures, which was absent in the response to pure photons. In this first study investigating gene expression in human blood cells exposed to mixed neutron-photon fields similar to an actual IND explosion, we have identified a number of genes responding to the 3 Gy dose that showed increasing expression as the neutron percentage increased. Such genes may serve as better indicators of the expected biological damage than a report of total physical dose, and thus provide more relevant information for treating physicians.

Project description:In the current global scenario there is always a possibility that a terrorist group might acquire material to produce a nuclear device. In the likelihood of a detonation of an improvised nuclear device (IND) the prompt radiation would consist of both photons (gamma rays) and neutrons. As neutrons generally have a high Relative Biological Effectiveness (RBE) for most physiological endpoints, it is important to understand the impact that neutrons would have on the biodosimetry methods that are being developed for medical triage purposes. We have previously reported transcriptomic response in human and mouse blood to neutron exposure and determined its RBE compared to photons for gene induction. In this report, we studied the effect of mixed neutron-photon radiation on gene expression in human peripheral blood in order to mimic an IND type radiation observed increase in number as well as expression level of genes with increasing percentage of neutron in mixed exposures which peaked at 25% neutron exposure which was maximum neutron percentage used for the study. We identified 156 genes that responded significantly to all the neutron-photon mixed exposures. These genes were found to be mainly involved in p53 signaling and DNA damage response. The 25% neutron exposure also showed maximum enrichment of known neutron associated gene ontology terms reported in the previous studies. Analysis of upstream regulators of gene expression showed strong activation of TP53, SLC29A1, PDCD1 and suppression of ANLN, AURK and ANXA2 across all exposures. While AGT was exclusively activated by neutron specific exposures. We confirmed the expression of genes which showed increased expression as a function of increased neutron percentage using quantitative real-time RT-PCR which matched with the microarray data. Thus, the data in this current study clearly indicates that gene expression can be used to estimate the percentage of the neutron component in mixed neutron-photon exposures.

Project description:The use of dose coefficients (DCs) based on the reference phantoms recommended by the International Commission on Radiological Protection (ICRP) with a fixed body size may produce errors to the estimated organ/tissue doses to be used, for example, for epidemiologic studies depending on the body size of cohort members. A set of percentile-specific computational phantoms that represent 10th, 50th, and 90th percentile standing heights and body masses in adult male and female Caucasian populations were recently developed by modifying the mesh-type ICRP reference computational phantoms (MRCPs). In the present study, these percentile-specific phantoms were used to calculate a comprehensive dataset of body-size-dependent DCs for photon external exposures by performing Monte Carlo dose calculations with the Geant4 code. The dataset includes the DCs of absorbed doses for 29 individual organs/tissues from 0.01 to 104 MeV photon energy, in the antero-posterior, postero-anterior, right lateral, left lateral, rotational, and isotropic geometries. The body-size-dependent DCs were compared with the DCs of the MRCPs in the reference body size, showing that the DCs of the MRCPs are generally similar to those of the 50th percentile standing height and body mass phantoms over the entire photon energy region except for low energies (≤ 0.03 MeV); the differences are mostly less than 10%. In contrast, there are significant differences in the DCs between the MRCPs and the 10th and 90th percentile standing height and body mass phantoms (i.e., H10M10 and H90M90). At energies of less than about 10 MeV, the MRCPs tended to under- and over-estimate the organ/tissue doses of the H10M10 and H90M90 phantoms, respectively. This tendency was revised at higher energies. The DCs of the percentile-specific phantoms were also compared with the previously published values of another phantom sets with similar body sizes, showing significant differences particularly at energies below about 0.1 MeV, which is mainly due to the different locations and depths of organs/tissues between the different phantom libraries. The DCs established in the present study should be useful to improve the dosimetric accuracy in the reconstructions of organ/tissue doses for individuals in risk assessment for epidemiologic investigations taking body sizes into account.

Project description:BACKGROUND:Radiation exposure due to the detonation of an improvised nuclear device remains a major security concern. Radiation from such a device involves a combination of photons and neutrons. Although photons will make the greater contribution to the total dose, neutrons will certainly have an impact on the severity of the exposure as they have high relative biological effectiveness. RESULTS:We investigated the gene expression signatures in the blood of mice exposed to 3 Gy x-rays, 0.75 Gy of neutrons, or to mixed field photon/neutron with the neutron fraction contributing 5, 15%, or 25% of a total 3 Gy radiation dose. Gene ontology and pathway analysis revealed that genes involved in protein ubiquitination pathways were significantly overrepresented in all radiation doses and qualities. On the other hand, eukaryotic initiation factor 2 (EIF2) signaling pathway was identified as one of the top 10 ranked canonical pathways in neutron, but not pure x-ray, exposures. In addition, the related mTOR and regulation of EIF4/p70S6K pathways were also significantly underrepresented in the exposures with a neutron component, but not in x-ray radiation. The majority of the changed genes in these pathways belonged to the ribosome biogenesis and translation machinery and included several translation initiation factors (e.g. Eif2ak4, Eif3f), as well as 40S and 60S ribosomal subunits (e.g. Rsp19, Rpl19, Rpl27). Many of the differentially downregulated ribosomal genes (e.g. RPS19, RPS28) have been causally associated with human bone marrow failure syndromes and hematologic malignancies. We also observed downregulation of transfer RNA processes, in the neutron-only exposure (p?<?0.005). Ingenuity Pathway Analysis (p?<?0.05) of differentially expressed genes predicted significantly suppressed activity of the upstream regulators c-Myc and Mycn, transcription factors known to control ribosome biogenesis. CONCLUSIONS:We describe the gene expression profile of mouse blood following exposure to mixed field neutron/photon irradiation. We have discovered that pathways related to protein translation are significantly underrepresented in the exposures containing a neutron component. Our results highlight the significance of neutron exposures that even the smallest percentage can have profound biological effects that will affect medical management and treatment decisions in case of a radiological emergency.

Dataset Information

Machine learning methodology for high throughput personalized neutron dose reconstruction in mixed neutron?+?photon exposures.

Publications

Machine learning methodology for high throughput personalized neutron dose reconstruction in mixed neutron + photon exposures.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets