Hello,
I am investigating proton-induced DNA damage using the molecularDNA example in Geant4 11.4.2. I am trying to understand a discrepancy between the current public implementation and the proton result reported by Shin et al.(2021).
My environment is:
- Geant4: 11.4.2
- OS: Ubuntu 24.04.4 LTS on WSL
- Compiler: g++ 13.3.0
- Threads: 4
For a direct comparison with Shin et al.(2021)., I first used the default human_cell geometry with 50 MeV protons and 10000 primaries.
The relevant damage and chemistry parameters were:
/process/dna/e-SolvationSubType Meesungnoen2002
/process/chem/TimeStepModel IRT_syn
/scheduler/endTime 5 ns/dnageom/radicalKillDistance 9 nm
/dnageom/interactionDirectRange 3.5 angstrom/dnadamage/directDamageLower 5 eV
/dnadamage/directDamageUpper 37.5 eV/dnadamage/indirectOHBaseChance 1.0
/dnadamage/indirectOHStrandChance 0.405
/dnadamage/inductionOHChance 0.0/dnadamage/indirectHBaseChance 1.0
/dnadamage/indirectHStrandChance 0.0
/dnadamage/inductionHChance 0.0/dnadamage/indirectEaqBaseChance 1.0
/dnadamage/indirectEaqStrandChance 0.0
/dnadamage/inductionEaqChance 0.0/gps/particle proton
/gps/energy 50 MeV
/run/beamOn 10000
I analysed the ROOT output using the human_cell.C analysis macro provided with molecularDNA.
I tested both physics options 2 and 4 while keeping the other parameters unchanged.
For the physics option 2, the result was:
Dose absorbed : 11.067 Gy
SSB total : 226.55 Gy^-1 Gbp^-1
DSB total : 7.67344 Gy^-1 Gbp^-1
SSB/DSB : 29.5239DSB direct : 0.719385
DSB indirect : 2.94807
DSB mixed : 0.324428
DSB hybrid : 3.68156
For physics option 4, the result was:
Dose absorbed : 11.0111 Gy
SSB total : 233.542 Gy^-1 Gbp^-1
DSB total : 9.10179 Gy^-1 Gbp^-1
SSB/DSB : 25.6589DSB direct : 1.10583
DSB indirect : 3.55849
DSB mixed : 0.595444
DSB hybrid : 3.84203
The change from physics option 4 to option 2 reduces the DSB yield from 9.10 to 7.67 Gy Gy^-1 Gbp^-1, but the option-2 result is still higher than I expected from the 50 MeV proton results reported by Shin et al.(2021).
Shin et al.(2021) report using G4EmDNAPhysics_option2 with the ELSEPA electron elastic scattering model and G4EmDNAChemistry_option3. Their damage parameters include Rdirect = 3.5 Å, thresholds of 5–37.5 eV, POH = 40.5%, radical kill distance = 9 nm, chemistry time = 5 ns, and dtmax = 0.5 ns.
My main question is therefore whether this magnitude of DNA damage yield is considered reasonable for the current Geant4 11.4.2 molecularDNA implementation, rather than whether the 2021 result must be reproduced exactly.
I also compared the 50 MeV result with my HTB-177 and MCF-7
cancer-cell simulations.
Each geometry was analysed using its corresponding ROOT analysis
macro and geometry-specific DNA/base-pair normalisation.
The human_cell and HTB-177 simulations shown below used 10,000 primary protons, whereas the MCF-7 result used 9,000 primaries. Therefore, this comparison is based on the normalised damage yields (Gy^-1 Gbp^-1), rather than on identical numbers of primary histories or raw numbers of DNA breaks.
| Geometry | Physics | Dose (Gy) | SSB | DSB | SSB/DSB |
|---|---|---|---|---|---|
| human_cell | p2 | 11.0670 | 226.550 | 7.673 | 29.52 |
| human_cell | p4 | 11.0111 | 233.542 | 9.102 | 25.66 |
| HTB-177 | p4 | 7.72783 | 234.035 | 8.98207 | 26.0558 |
| MCF-7 | p4 | 10.2164 | 225.934 | 8.303 | 27.21 |
For physics option 4, the three cell geometries therefore give normalised DSB yields of a similar magnitude, approximately 8.3–9.1 Gy^-1 Gbp^-1, despite differences in cell geometry and, for MCF-7, the number of primary histories.
I found two previous Geant4 Forum discussion that appear closely related to this issue :
In the 2022 discussion, another user reported higher SSB and DSB yields when attempting to reproduce Shin et al.(2021) proton results. The discussion noted that the `molecularDNA` example was using the newer IRT-syn chemistry implementation, and differences with the implementation used in the publication were being investigated.
A similar discrepancy was reported again in the 2023 discussion using the human_cell geometry and a closely related parameter configuration.
Because I am now using Geant4 11.4.2, I would like to understand whether this behaviour is still expected in the current implementation.
I would therefore appreciate clarification on the following points:
- Are SSB and DSB yields of this magnitude considered reasonable for 50 MeV proton irradiation with the current molecularDNA implementation in Geant4 11.4.2? In particular, is a DSB yield of approximately 7.67 Gy^-1 Gbp^-1 with `-p 2` and 9.10 Gy^-1 Gbp^-1 with `-p 4` within the expected range for the current implementation?
- The previous 2022 and 2023 Geant4 Forum discussions reported a similar tendency for molecularDNA to give higher SSB/DSB yields than the published Shin et al. results. Is this discrepancy still expected with Geant4 11.4.2, or should the current molecularDNA implementation now be able to reproduce the older published result more closely?
- Does the current `-p 2` configuration correspond sufficiently closely to the G4EmDNAPhysics_option2 + ELSEPA configuration used by Shin et al. (2021)? Also, should differences be expected between the current `IRT_syn` chemistry implementation and the chemistry/IRT implementation used in that publication?
- Using the corresponding ROOT analysis macro and DNA normalisation for each geometry, I obtain p4 DSB yields of approximately 9.10 Gy^-1 Gbp^-1 for human_cell, 8.98 Gy^-1 Gbp^-1 for HTB-177, and 8.30 Gy^-1 Gbp^-1 for MCF-7. Is this similarity in the normalised DSB yield across the three geometries reasonable for the current molecularDNA implementation under the same 50 MeV proton irradiation and damage-model parameters?
I have not modified the Geant4 or molecularDNA source code. Thank you for any clarification.