biospectral

Predict absorbance spectra for a set of SMILES using Gaussian 16. Running the TD-DFT predictions requires a Gaussian 16 license and installation.

To run the scripts that generate the Gaussian 16 inputs and call the TD-DFT prediction code, scripts are provided to submit jobs via various job scheduler interfaces (e.g., SLURM, and a system unique to MIT Lincoln Lab's Supercloud cluster). The underlying pipeline is the same.

Note that the TD-DFT simulations take a long time to run and can take months to run for 1000s of molecules on multiple CPUs.

Step 1A. Run through SLURM in batched mode

A1. Generate Gaussian 16 input files

db_to_gaussian_input.sh
Parameters:
--smiles-file: string CSV file containing input SMILES
--name-col: string name of column in smiles-file containing name (unique IDs) of molcules used to label output files
--smiles-col: string name of column in smiles-file containing SMILES
-d, --dir: string directory to store generated 3D structure files and Gaussian input files
-f, --functional: string of Gaussian 16 functional to be used in TD-DFT calculation
-b, --basis: string of Gaussian 16 basis set to be used in TD-DFT calculation
--mem: int memory to be allocated for TD-DFT jobs
--cpus: int cpus to be allocated for TD-DFT jobs

A2. Predict absorbance spectra using TD-DFT with Gaussian 16

tddft_batched.sh
Parameters:
--input-dir: string directory containing Gaussian 16 input files
--output-dir: string directory for saving output files
--jobs: int number of separate Slurm jobs to split the work into

Step 1B. Run single jobs

(can be run in batch on MIT Lincoln Lab Supercloud cluster using LLMapReduce --mapper mapper.sh --input <path_to_raw_data> --output output/ --np <cpu usage specification>)

predict_optical_properties.sh
Parameters:
-i: string Path to directory for saving generated 3D molecular coordinates files
-n: string Name of input molecule (used as label for saving files)
-s: string SMILES string of input molecule
-o: string Path to directory for saving TD-DFT prediction outputs
-m: int Memory to allocate for Gaussian 16 TD-DFT job
-c: int CPU number to allocate for Gaussian 16 TD-DFT job
-f: string Name of TD-DFT functional to use for Gaussian 16 TD-DFT job.\

Step 2.5 Retrieve singlet excitation values from Gaussian

run grep Singlet *log in the output directory of Gaussian predictions
TODO: incorporate this with the rest of the workflow

Step 3. Generate UV-Vis absorbance spectra from TD-DFT Singlets

get_spectra.py
Parameters:
--singlets: string file name with excitation information from TD-DFT predictions (generated in step 2.5)
--out-file: string name for CSV file in which spectra are saved
--lower-bound: int lower bound (in nm) for the wavelengths for spectrum generation
--upper-bound: int upper bound (in nm) for the wavelengths for spectrum generation
-s: float parameter determining gaussian FWHM during convolution
--fixed-width: if flag is set, all peaks have equal width with FWHM determined b s in nm.\