artemponomarevjetski/nasa-csharp-reinforced-machine-learning-4-space-particles-induced-chromosome-aberrations

Selected NASA work

SupervisedMachineLearningForChromosomalAberrations

This is a compendium of my best work for NASA. All these projects evolved on the span of at least 2 decades. Out of a huge heap of publications and code developed on the subject of space radiation and its effect on human health, I included herein my work closer to AI/ML/DS areas, in which area I continue my career now.

For those who like to ask about my background (perhaps, recruiters and employers), the flowchart of my career is as follows: general high aptitude in science and math in high school -> olympiads in math and physics -> boarding school for gifted children in math and physics -> undergraduate studies in a physics department in 2 places -> a student in a theoretical physics group led by a Nobel Prize winner, Vitaly L. Ginzburg -> Ph.D. in applied physics from Columbia University -> internship in Los Alamos National Lab -> post-doc at math department at UC Berkeley -> post-doc at NASA JSC -> assistant professor at BCM, Houston -> senior research scientist at NASA JSC -> senior data scientist at FORA capital -> doing projects from home during COVID-19 pandemic -> next stage? I am interested in going deeper into ML/DS/AI, especially NLP algorithms; I want to work at a company where I can apply my best skills...

So, my work at NASA JSC, that goes into this repo, has the following flowchart: particle physics of relativistic ions in space and radiation transport through matter -> human DNA damage and breakage patterns from space particles -> superposition of radiation tracks and coiled chromatin translates into a unique high-energy DNA breakage pattern -> DNA breakage leads to chomosome aberrations with complex aberrations being the signature of space particles -> due to combinatorial complexity of chromosome aberratation, a supervised training algorithm in employed to match the algorithmic classification of aberrations to a given classification scheme employed by a particular lab (or labs) -> the frequency and type of aberrations has different sensitivity to different energies, which suggest a new strategy for spacecraft radiation shielding -- only the part of particle spectrum where DNA is most sensitive needs to be suppressed -> particle spectrum can be manipulated by different materials with low atomic numbers and high atomic numbers -> a compound radiation shield is proposed with hydrogen, water, alumunum and hafnium -> hafnium has additional pathways of energy absorption -- high neutron absortion ability and, possibly, the isomer spin resonance -> the overall optimization demands a "super" machine learning approach -- with a multi-parametric feature space -> the proposed feature space has the following categorical and numeric axes: target -- human or electroncis; mission stages (LEO, GEO, Lagrange points, deep space, planetory surfaces with and w/o atmospheres, access to in situ materials); habitality considerations -- astronaut life regimen and intra-habitat radiation environment; antioxydant diets and human body "radiation hardening" with antioxydant drugs -> real-time interation between the developed AI system and the flight surgeon during the mission -> in-flight and on-board mitigation... TBD.