GitHub - ALH477/science-flake: Industry-standard development environments for: AI/ML Training, Reinforcement Learning, Scientific Computing, Biology & Genomics, and Structural Biology.

╔══════════════════════════════════════════════════════════════════════════════╗
║                                                                              ║
║   ███████╗ ██████╗██╗███████╗███╗   ██╗ ██████╗███████╗                      ║
║   ██╔════╝██╔════╝██║██╔════╝████╗  ██║██╔════╝██╔════╝                      ║
║   ███████╗██║     ██║█████╗  ██╔██╗ ██║██║     █████╗                        ║
║   ╚════██║██║     ██║██╔══╝  ██║╚██╗██║██║     ██╔══╝                        ║
║   ███████║╚██████╗██║███████╗██║ ╚████║╚██████╗███████╗                      ║
║   ╚══════╝ ╚═════╝╚═╝╚══════╝╚═╝  ╚═══╝ ╚═════╝╚══════╝                      ║
║                                                                              ║
║   ███████╗██╗      █████╗ ██╗  ██╗███████╗                                   ║
║   ██╔════╝██║     ██╔══██╗██║ ██╔╝██╔════╝                                   ║
║   █████╗  ██║     ███████║█████╔╝ █████╗                                     ║
║   ██╔══╝  ██║     ██╔══██║██╔═██╗ ██╔══╝                                     ║
║   ██║     ███████╗██║  ██║██║  ██╗███████╗                                   ║
║   ╚═╝     ╚══════╝╚═╝  ╚═╝╚═╝  ╚═╝╚══════╝                                   ║
║                                                                              ║
║   Production-Grade Scientific Computing & AI/ML Suite for NixOS             ║
║   ─────────────────────────────────────────────────────────────              ║
║   DeMoD LLC                                                                  ║
║                                                                              ║
╚══════════════════════════════════════════════════════════════════════════════╝

Overview

Industry-standard development environments for:

Domain	Templates	Key Tools
AI/ML Training	`ml-training`, `llm`, `vision`, `audio-ml`	PyTorch, Lightning, DeepSpeed, PEFT
Reinforcement Learning	`pufferlib`, `rl`	PufferLib (1M+ steps/s), SB3, MuJoCo
Scientific Computing	`python`, `julia`, `r`	NumPy, SciPy, Tidyverse
Biology & Genomics	`bioinformatics`, `crispr`, `singlecell`	scanpy, BLAST, CRISPR tools, Bioconductor
Structural Biology	`protein`, `molecular-dynamics`	PyMOL, AlphaFold tools, GROMACS
System Tools	NixOS module	QGIS, PyMOL, LaTeX, Ollama

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃  SYSTEM LEVEL                      PROJECT LEVEL                          ┃
┃  ─────────────                     ─────────────                          ┃
┃  ┌─────────────────────┐           ┌─────────────────────────────────┐    ┃
┃  │ programs.science    │           │ nix flake init -t ...#llm       │    ┃
┃  │   .enable = true    │           │                                 │    ┃
┃  │   .cuda.enable      │           │ ┌─────────┐  ┌─────────┐        │    ┃
┃  │   .biology.crispr   │           │ │ llm/    │  │ crispr/ │  ...   │    ┃
┃  │                     │           │ │ PyTorch │  │ Primer3 │        │    ┃
┃  │ Installs:           │           │ │ vLLM    │  │ BLAST   │        │    ┃
┃  │ - llama-cpp         │           │ │ TRL     │  │ scanpy  │        │    ┃
┃  │ - QGIS, PyMOL       │           │ └─────────┘  └─────────┘        │    ┃
┃  │ - LaTeX, IGV        │           │                                 │    ┃
┃  └─────────────────────┘           │ Isolated • Reproducible • GPU   │    ┃
┃                                    └─────────────────────────────────┘    ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛

Quick Start

1. Add to Flake Inputs

{
  inputs = {
    nixpkgs.url = "github:nixos/nixpkgs/nixos-24.11";
    science.url = "path:/home/user/science-flake";
    science.inputs.nixpkgs.follows = "nixpkgs";
  };
}

2. Enable System Module

nixosConfigurations.myMachine = nixpkgs.lib.nixosSystem {
  modules = [
    science.nixosModules.science
    {
      programs.science = {
        enable = true;
        cuda.enable = true;
        bleedingEdge = {
          enable = true;
          llm = true;         # llama-cpp
        };
        rl.enable = true;     # MuJoCo dependencies
      };
    }
  ];
};

3. Create Project Environments

# LLM Fine-tuning
mkdir llm-project && cd llm-project
nix flake init -t /path/to/science-flake#llm
direnv allow

# CRISPR Guide Design
mkdir crispr-project && cd crispr-project
nix flake init -t /path/to/science-flake#crispr
direnv allow

# Single-Cell Analysis
mkdir singlecell-project && cd singlecell-project
nix flake init -t /path/to/science-flake#singlecell
direnv allow

4. Quick Shell Access

# Direct shell access
nix-shell shell.nix                                    # Core only
nix-shell shell.nix --arg modules '[ "core" "bio" ]'   # Multiple modules

# Interactive TUI selector (recommended)
python tui.py

# Commands available in shell:
sci-help      # Quick reference for active modules
sci-modules   # List all available modules  
sci-jupyter   # Launch JupyterLab
sci-tui       # Re-launch interactive selector

5. TUI Navigation

┏━━━ SCIENCE FLAKE ━━━┓
   Module Selector & Guide

▸ Core
  [✓] Core Tools         Essential data processing...
▸ Biology  
  [ ] Bioinformatics     NGS, alignment, variant calling
  [✓] CRISPR/Gene Edit   Guide design, off-target analysis
  ...

Selected (2): Core Tools, CRISPR/Gene Editing

↑↓/jk:navigate  Space:toggle  d/→:details  a:all  n:none  r:run  q:quit

Keys:

↑↓ or j/k — Navigate modules
Space or Enter — Toggle selection
d or → — View module details & examples
a — Select all modules
n — Reset to core only
r — Generate run command
q — Quit

Templates Reference

AI/ML Model Building

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃  TEMPLATE         DESCRIPTION                          KEY PACKAGES        ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┫
┃  ml-training      Full training stack                  PyTorch, Lightning  ┃
┃                   Distributed, mixed precision         DeepSpeed*, NCCL    ┃
┃                                                                            ┃
┃  llm              LLM training & fine-tuning           transformers, PEFT* ┃
┃                   RLHF, DPO, inference                 TRL*, vLLM*, Ollama ┃
┃                                                                            ┃
┃  vision           Detection, segmentation, gen         torchvision, YOLO*  ┃
┃                   Multimodal                           SAM*, Diffusers*    ┃
┃                                                                            ┃
┃  audio-ml         Audio/speech/music ML                torchaudio, librosa ┃
┃                   DSP, neural synthesis                Faust, audiocraft*  ┃
┃                                                                            ┃
┃                                                   * = pip install required ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛

Biology & Genomics

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃  TEMPLATE         DESCRIPTION                          KEY PACKAGES        ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┫
┃  bioinformatics   Core genomics pipeline               BLAST, samtools     ┃
┃                   Alignment, variant calling           bcftools, bedtools  ┃
┃                                                                            ┃
┃  crispr           CRISPR/gene editing design           Primer3, BLAST      ┃
┃                   Guide RNA, off-target analysis       CRISPResso2*, IGV   ┃
┃                                                                            ┃
┃  singlecell       Single-cell & spatial omics          scanpy, anndata     ┃
┃                   Trajectory, integration              scvelo, squidpy     ┃
┃                                                                            ┃
┃  rnaseq           RNA-seq quantification               STAR, salmon        ┃
┃                   Differential expression              DESeq2, kallisto    ┃
┃                                                                            ┃
┃  protein          Structural biology                   PyMOL, MDAnalysis   ┃
┃                   Protein engineering                  ESMFold*, OpenFold* ┃
┃                                                                            ┃
┃  synbio           Synthetic biology workflows          COBRApy, ViennaRNA  ┃
┃                   Metabolic engineering                EMBOSS, Primer3     ┃
┃                                                                            ┃
┃  metagenomics     Microbiome analysis                  Kraken2, MetaPhlAn  ┃
┃                   Taxonomic profiling                  MEGAHIT, prokka     ┃
┃                                                                            ┃
┃  phylogenetics    Evolutionary analysis                RAxML, IQ-TREE      ┃
┃                   Tree building & visualization        MrBayes, FigTree    ┃
┃                                                                            ┃
┃                                                   * = pip install required ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛

Reinforcement Learning

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃  TEMPLATE         DESCRIPTION                          KEY PACKAGES        ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┫
┃  pufferlib        High-perf RL (1M+ steps/sec)         PufferLib*, CleanRL ┃
┃                   Ocean, Atari, NetHack, NMMO          Gymnasium, PyTorch  ┃
┃                                                                            ┃
┃  rl               General RL research                  SB3, PettingZoo     ┃
┃                   Multi-agent, MuJoCo                  MuJoCo, Ray RLlib*  ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛

Scientific Computing

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃  TEMPLATE         DESCRIPTION                          KEY PACKAGES        ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┫
┃  python           Data science standard                NumPy, Pandas       ┃
┃                                                        scikit-learn, Jupyter┃
┃                                                                            ┃
┃  python-gpu       GPU-accelerated ML                   PyTorch+CUDA, JAX   ┃
┃                                                                            ┃
┃  julia            Scientific Julia                     Plots, DataFrames   ┃
┃                                                        DifferentialEquations┃
┃                                                                            ┃
┃  r                Statistics                           Tidyverse, ggplot2  ┃
┃                                                        Shiny, RStudio      ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛

Module Options

Core

programs.science = {
  enable = true;
  
  # GPU Acceleration
  cuda = {
    enable = true;
    cudaCapabilities = [ "8.6" ];  # RTX 30xx
  };
  rocm.enable = false;  # AMD GPU (mutually exclusive with CUDA)
  
  # Core tools (always recommended)
  core.enable = true;  # pandoc, typst, gnuplot
};

Biology & Genomics (NEW)

programs.science = {
  biology = {
    enable = true;           # Core bio tools (seqkit, emboss)
    
    # Workflow modules
    genomics.enable = true;        # NGS: BLAST, BWA, samtools, bcftools
    transcriptomics.enable = true; # RNA-seq: STAR, salmon, kallisto
    singleCell.enable = true;      # scanpy, anndata, scvelo, squidpy
    crispr.enable = true;          # Primer3, guide design, off-target
    structural.enable = true;      # PyMOL, Chimera, MDAnalysis
    phylogenetics.enable = true;   # RAxML, IQ-TREE, MrBayes
    metagenomics.enable = true;    # Kraken2, MetaPhlAn, MEGAHIT
    synbio.enable = true;          # COBRApy, ViennaRNA
    molecularDynamics.enable = true; # GROMACS (CUDA if enabled)
    
    # GUI applications
    gui.enable = true;       # PyMOL, IGV, Fiji, JalView
    
    # R Bioconductor
    bioconductor.enable = true;  # DESeq2, Seurat, GenomicRanges
  };
};

Traditional Science

programs.science = {
  latex = {
    enable = true;
    scheme = "medium";  # minimal | small | medium | full
  };
  chemistry.enable = true;    # Avogadro, GROMACS
  physics.enable = true;      # Stellarium, ROOT
  geospatial.enable = true;   # QGIS, GDAL
  mathematics.enable = true;  # Maxima, Octave
  electronics.enable = true;  # KiCad, GNU Radio
};

Bleeding Edge AI/ML

programs.science = {
  bleedingEdge = {
    enable = true;
    llm = true;           # llama-cpp, Ollama
    mlops = true;         # dvc, mlflow
    visualization = true; # manim
  };
  
  rl = {
    enable = true;
    simulation = true;    # MuJoCo dependencies
  };
};

Biology Workflows

CRISPR Guide Design & Analysis

cd crispr-project

# Environment includes: Primer3, BLAST, BWA, samtools, IGV

# Install specialized CRISPR tools
pip install crispresso2        # Editing analysis
pip install cas-offinder       # Off-target prediction

# Guide design workflow
python -c "
from Bio import SeqIO
from Bio.Seq import Seq
import primer3

# Load target sequence
target = 'ATGCGATCGATCGATCGATCGATCG'

# Find PAM sites (NGG for SpCas9)
pam_sites = [i for i in range(len(target)-2) 
             if target[i+1:i+3] == 'GG']

# Design primers for validation PCR
primers = primer3.design_primers(
    seq_args={'SEQUENCE_TEMPLATE': target},
    global_args={'PRIMER_PRODUCT_SIZE_RANGE': [[100, 300]]}
)
"

# Off-target analysis with BLAST
blastn -query guides.fa -db genome_db -outfmt 6 -max_target_seqs 100

# Visualize in IGV
igv -g hg38 aligned_reads.bam

Single-Cell RNA-seq Analysis

cd singlecell-project

# Environment includes: scanpy, anndata, scvelo, squidpy

# Install additional tools
pip install scvi-tools         # Deep learning
pip install celltypist         # Annotation

python -c "
import scanpy as sc
import anndata as ad

# Load 10x data
adata = sc.read_10x_mtx('filtered_feature_bc_matrix/')

# Standard preprocessing
sc.pp.filter_cells(adata, min_genes=200)
sc.pp.filter_genes(adata, min_cells=3)
sc.pp.normalize_total(adata, target_sum=1e4)
sc.pp.log1p(adata)
sc.pp.highly_variable_genes(adata, n_top_genes=2000)

# Dimensionality reduction
sc.tl.pca(adata)
sc.pp.neighbors(adata)
sc.tl.umap(adata)
sc.tl.leiden(adata)

# Visualization
sc.pl.umap(adata, color='leiden')
"

RNA-seq Differential Expression

cd rnaseq-project

# Alignment with STAR
STAR --genomeDir star_index \
     --readFilesIn reads_R1.fq.gz reads_R2.fq.gz \
     --readFilesCommand zcat \
     --outSAMtype BAM SortedByCoordinate

# Quantification with salmon
salmon quant -i salmon_index -l A \
       -1 reads_R1.fq.gz -2 reads_R2.fq.gz \
       -o quants/sample1

# R analysis with DESeq2
Rscript -e "
library(DESeq2)
library(tximport)

# Import salmon counts
files <- list.files('quants', pattern='quant.sf', recursive=TRUE, full.names=TRUE)
txi <- tximport(files, type='salmon', tx2gene=tx2gene)

# DESeq2 analysis
dds <- DESeqDataSetFromTximport(txi, colData, ~condition)
dds <- DESeq(dds)
res <- results(dds, contrast=c('condition', 'treated', 'control'))
"

Protein Structure Prediction

cd protein-project

# Install bleeding-edge protein ML
pip install esm                # Meta's ESM
pip install openfold           # OpenFold (requires setup)

# ESMFold prediction
python -c "
import torch
import esm

# Load ESMFold
model = esm.pretrained.esmfold_v1()
model = model.eval().cuda()

# Predict structure
sequence = 'MKTVRQERLKSIVRILERSKEPVSGAQLAEELSVSRQVIVQDIAYLRSLGYNIVATPRGYVLAGG'
with torch.no_grad():
    output = model.infer_pdb(sequence)

# Save PDB
with open('predicted.pdb', 'w') as f:
    f.write(output)
"

# Visualize with PyMOL
pymol predicted.pdb

Molecular Dynamics with GROMACS

cd md-project

# Prepare system
gmx pdb2gmx -f protein.pdb -o protein.gro -water spce
gmx editconf -f protein.gro -o box.gro -c -d 1.0 -bt cubic
gmx solvate -cp box.gro -cs spc216.gro -o solvated.gro -p topol.top

# Add ions
gmx grompp -f ions.mdp -c solvated.gro -p topol.top -o ions.tpr
gmx genion -s ions.tpr -o system.gro -p topol.top -pname NA -nname CL -neutral

# Energy minimization
gmx grompp -f em.mdp -c system.gro -p topol.top -o em.tpr
gmx mdrun -v -deffnm em

# Production MD (GPU accelerated if CUDA enabled)
gmx grompp -f md.mdp -c npt.gro -t npt.cpt -p topol.top -o md.tpr
gmx mdrun -deffnm md -nb gpu

Directory Structure

science-flake/
├── flake.nix                 # Root flake with module exports
├── shell.nix                 # Unified dev shell with module selection
├── tui.py                    # Interactive curses TUI selector
├── README.md                 # This file
│
├── modules/
│   ├── science.nix           # NixOS module (system-level)
│   └── science-hm.nix        # Home Manager module
│
├── overlays/
│   ├── science.nix           # Package customizations
│   ├── biology.nix           # Biology-specific overlays (NEW)
│   └── bleeding-edge.nix     # Latest AI/ML packages
│
└── templates/
    ├── python/               # Standard data science
    ├── python-gpu/           # GPU-accelerated ML
    ├── julia/                # Scientific Julia
    ├── r/                    # R statistics
    │
    ├── ml-training/          # Full ML training stack
    ├── llm/                  # LLM development
    ├── vision/               # Computer vision
    ├── audio-ml/             # Audio/DSP ML
    │
    ├── pufferlib/            # High-perf RL
    ├── rl/                   # General RL research
    │
    ├── bioinformatics/       # Core genomics
    ├── crispr/               # CRISPR/gene editing (NEW)
    ├── singlecell/           # Single-cell omics (NEW)
    ├── rnaseq/               # RNA-seq (NEW)
    ├── protein/              # Structural biology (NEW)
    ├── synbio/               # Synthetic biology (NEW)
    ├── metagenomics/         # Microbiome (NEW)
    └── phylogenetics/        # Evolutionary (NEW)

GPU Configuration

NVIDIA CUDA

# In your NixOS configuration
hardware.nvidia = {
  package = config.boot.kernelPackages.nvidiaPackages.stable;
  modesetting.enable = true;
  open = false;  # or true for open-source driver (RTX 20xx+)
};

programs.science.cuda = {
  enable = true;
  cudaCapabilities = [ "8.6" ];  # Match your GPU
};

GPU Series	Capability
RTX 20xx	7.5
RTX 30xx	8.6
RTX 40xx	8.9
H100	9.0

AMD ROCm

hardware.amdgpu.opencl.enable = true;
programs.science.rocm.enable = true;

Overlays

nixpkgs.overlays = [
  science.overlays.default       # Package customizations
  science.overlays.biology       # Biology-specific (NEW)
  science.overlays.bleedingEdge  # Latest AI/ML
  science.overlays.unstable      # nixpkgs-unstable access
];

Available packages from overlays:

ML/AI:

python-ml-bleeding - Full ML stack
python-rl-bleeding - RL research stack
llama-cpp-cuda - GPU-accelerated inference

Biology:

python-crispr - CRISPR design & analysis
python-singlecell - Single-cell genomics
python-structural - Protein structure
python-sysbio - Systems biology
python-bioml-bleeding - ML for biology
r-bioconductor - Bioconductor packages
r-crispr - CRISPR R packages
bio-alignment-suite - CLI alignment tools
bio-ngs-suite - NGS processing
bio-crispr-suite - CRISPR CLI tools
bio-phylo-suite - Phylogenetics
synbio-tools - Synthetic biology

Chemistry/MD:

gromacsCuda - GPU-accelerated molecular dynamics
gromacs-cuda-full - Full GROMACS + CUDA + MPI

License

MIT

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃                                                                            ┃
┃   DeMoD LLC                                                                ┃
┃   "Build different."                                                       ┃
┃                                                                            ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛

Name		Name	Last commit message	Last commit date
Latest commit History 2 Commits
LICENSE		LICENSE
README.md		README.md
biology.nix		biology.nix
bleeding-edge.nix		bleeding-edge.nix
science-hm.nix		science-hm.nix
science.nix		science.nix
shell.nix		shell.nix
tui.py		tui.py

Folders and files

Latest commit

History

Repository files navigation

Overview

Quick Start

1. Add to Flake Inputs

2. Enable System Module

3. Create Project Environments

4. Quick Shell Access

5. TUI Navigation

Templates Reference

AI/ML Model Building

Biology & Genomics

Reinforcement Learning

Scientific Computing

Module Options

Core

Biology & Genomics (NEW)

Traditional Science

Bleeding Edge AI/ML

Biology Workflows

CRISPR Guide Design & Analysis

Single-Cell RNA-seq Analysis

RNA-seq Differential Expression

Protein Structure Prediction

Molecular Dynamics with GROMACS

Directory Structure

GPU Configuration

NVIDIA CUDA

AMD ROCm

Overlays

License

About

Topics

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages