AI-Driven Drug Discovery & Diagnostics

Deep learning for
molecular design, protein
engineering, and clinical diagnostics.

Dynotx trains graph neural networks, protein language models, and diffusion-based generative models on GPU clusters to discover novel therapeutics — from in silico screening to preclinical candidates, compressed from years to months.

2.4M  molecules screened
89.2%  AUROC (MoleculeNet)
3  preclinical candidates
12K  GPU-hrs / month
SCROLL
01 / Technology

Three architectures.
One discovery pipeline.

Our platform integrates graph neural networks for molecular property prediction, transformer-based protein language models, and diffusion models for de novo molecular generation. Each model is trained on large-scale biochemical datasets and deployed through a unified inference stack optimized for NVIDIA GPU architecture.

FIG. 1 — PLATFORM ARCHITECTURE
v2.4
Molecular Input SMILES · SDF · 2.4M ZINC + ChEMBL
Graph Representation Atom-bond graph · 9 node feats · 5 edge feats
GNN Encoder (MolGNN) 8 message-passing layers · d=512
Protein Encoder (ProtLM) ESM-2 backbone · 650M params · d=1280
Diffusion Decoder (Diffuse) 200 denoising steps · 3D equivariant
Candidate Molecules Docked · scored · ADME-filtered

Architecture Overview

The pipeline begins with molecular inputs from the ZINC and ChEMBL databases — 2.4 million drug-like compounds represented as atom-bond graphs with 9-dimensional node features (atomic number, charge, hybridization, aromaticity) and 5-dimensional edge features (bond type, ring membership, conjugation).

A graph neural network (Dynotx-MolGNN) performs message passing over the molecular graph to predict properties: binding affinity, toxicity, solubility, and bioavailability. In parallel, a protein language model (Dynotx-ProtLM, based on the ESM-2 architecture) encodes target protein sequences into structural and functional embeddings.

The diffusion model (Dynotx-Diffuse) generates novel 3D molecular structures conditioned on the target protein binding pocket. Generated candidates are docked, scored, and filtered through ADME prediction — producing ranked drug candidates ready for wet-lab validation.

training_data ZINC15 (2.4M) + ChEMBL (1.9M) + PDB (180K structures)
framework PyTorch 2.4 · PyG · CUDA 12.5 · cuDNN 9
precision bf16 mixed precision · FSDP sharding
inference TensorRT · ONNX export · < 50ms / molecule
deployment Triton Inference Server · H100 SXM
02 / Pipeline

From molecular database
to clinical candidate.

Our end-to-end pipeline processes molecular datasets through GPU-accelerated training, TensorRT-optimized inference, and clinical deployment. Each stage is designed for scale — 2.4M molecules screened, 12,000 GPU-hours per month, 3 drug candidates in preclinical evaluation.

STEP 01

Data Curation

Molecular datasets from ZINC15, ChEMBL, and Protein Data Bank. Standardized SMILES, 3D conformers generated via RDKit + OpenMM. Protein structures prepared with AlphaFold2 predictions and experimental PDB data.

ZINC15 · ChEMBL32 · PDB · ~4.3M compounds
STEP 02

GPU Training

Distributed training on H100 SXM clusters. FSDP sharding across 8 GPUs, bf16 mixed precision, gradient checkpointing. Separate training runs for each model architecture with shared data pipeline.

PyTorch 2.4 · CUDA 12.5 · NCCL · bf16
STEP 03

Inference & Screening

Models exported to TensorRT and served via Triton. Virtual screening of 2.4M compounds against target proteins. Diffusion model generates novel candidates conditioned on binding pocket geometry.

TensorRT · Triton · < 50ms / mol inference
STEP 04

Clinical Deployment

Top candidates validated through ADME prediction, toxicity filtering, and docking score ranking. Promising hits advance to wet-lab validation. 3 candidates currently in preclinical evaluation across oncology and infectious disease targets.

3 preclinical candidates · oncology + ID
train_protlm.yaml
YAML
# Dynotx-ProtLM v2.4 — distributed training configuration
# Cluster: 8× NVIDIA H100 80GB SXM · NVLink · InfiniBand

model:
  architecture: esm2_transformer
  params: 650_000_000
  layers: 33
  hidden_dim: 1280
  attention_heads: 20
  max_sequence_len: 1024
  positional_encoding: rotary

training:
  gpu_cluster: 8x H100 SXM 80GB
  strategy: FSDP  # Fully Sharded Data Parallel
  batch_size: 64  # per-GPU, grad accumulation = 4
  global_batch: 2048
  learning_rate: 4.0e-4
  warmup_steps: 8000
  total_steps: 250_000
  gradient_checkpointing: true
  mixed_precision: bf16
  optimizer: AdamW
  weight_decay: 0.01
  scheduler: cosine_decay

data:
  source: UniRef50 + Pfam + internal sequences
  total_sequences: 45_000_000
  masking: span_masking  # 15% masked, avg span 3
  augmentation: sequence_crop, random_truncation

compute:
  estimated_gpu_hours: 14_400
  peak_memory_per_gpu: 68GB  # FSDP sharded
  interconnect: NVLink 4.0 + InfiniBand HDR
  framework: PyTorch 2.4 · PyG · CUDA 12.5

inference:
  runtime: tensorrt
  server: triton_inference_server
  quantization: fp16
  latency: 47ms  # per protein, 512 AA avg
  throughput: 1200 seq/s
03 / Models

Three models powering
the discovery stack.

Each model addresses a distinct stage of the drug discovery pipeline — molecular property prediction, protein understanding, and de novo drug design. All models are trained, validated, and deployed independently, then composed into a unified screening pipeline.

Model Architecture Parameters Training Performance
Dynotx-MolGNN
dynotx-molgnn-v2.1
Graph neural network for molecular property prediction. Predicts binding affinity, toxicity, solubility, and bioactivity from atom-bond graph representations. 8 message-passing layers with residual connections and edge-conditioned convolutions.
GNN (DMPNN) 8 message-passing layers
d=512, 9 node feats
5 edge feats
edge-conditioned conv
12.4M FP16 inference
ONNX export
PyG backend
2,800 GPU-hrs 8× H100 SXM
batch=512
bf16 mixed precision
89.2%
AUROC (MoleculeNet avg)
Dynotx-ProtLM
dynotx-protlm-v2.4
Protein language model based on ESM-2 architecture. Learns structural and functional representations from 45M protein sequences. Used for target identification, binding site prediction, and protein function annotation across the drug discovery pipeline.
ESM-2 Transformer 33 layers · 20 heads
d=1280
rotary positional encoding
span masking (15%)
650M FP16 inference
Triton serving
FSDP training
14,400 GPU-hrs 8× H100 SXM
global batch=2048
250K training steps
0.83
Contact prediction F1
Dynotx-Diffuse
dynotx-diffuse-v1.3
3D equivariant diffusion model for de novo molecular generation. Generates novel drug-like molecules conditioned on target protein binding pocket geometry. 200 denoising steps with SE(3)-equivariant message passing for 3D coordinate generation.
SE(3) Diffusion 200 denoising steps
3D equivariant GNN
pocket-conditioned
3D coordinate output
85M FP16 inference
batch=64
ONNX export
6,400 GPU-hrs 8× H100 SXM
batch=64
classifier-free guidance
71.4%
Docking success rate
04 / Infrastructure

GPU training pipeline
built for scale.

Our models are trained on NVIDIA H100 SXM clusters with FSDP sharding, bf16 mixed precision, and gradient checkpointing. The full stack — from data curation to TensorRT inference — runs on our internal compute infrastructure. We are scaling to 12,000 GPU-hours per month and need additional compute capacity to support expanded molecular screening and protein engineering campaigns.

01
Molecular Data Pipeline
Compounds from ZINC15 (2.4M drug-like molecules) and ChEMBL32 (1.9M bioactive compounds) are standardized, filtered by Lipinski rules, and converted to 3D conformers. Protein structures from PDB (180K experimental) supplemented with AlphaFold2 predictions for targets without crystal structures.
RDKit · OpenMM · AlphaFold2 · ~4.3M compounds · 2.1TB processed
02
Distributed GPU Training
Training runs on 8× H100 SXM 80GB with FSDP sharding, bf16 mixed precision, and gradient checkpointing. ProtLM requires 68GB peak memory per GPU (FSDP sharded). NCCL over NVLink + InfiniBand for all-reduce. Separate training scripts per model architecture, shared data loader.
PyTorch 2.4 · FSDP · NCCL · bf16 · 8× H100 SXM 80GB
03
TensorRT Inference & Screening
Models exported to ONNX and compiled with TensorRT for FP16 inference. Served via Triton Inference Server on H100 SXM. Virtual screening of 2.4M compounds completes in under 48 hours. Diffusion model generates 50K novel candidates per target pocket per run.
TensorRT · Triton · FP16 · < 50ms / molecule · 2.4M screened
04
Clinical Candidate Selection
Top candidates filtered through ADME prediction (solubility, permeability, metabolic stability), toxicity prediction (hERG, hepatotoxicity, mutagenicity), and docking score ranking. 3 candidates currently in preclinical evaluation: DTX-201 (oncology), DTX-114 (antiviral), DTX-307 (antibacterial).
ADME · tox filter · docking · 3 preclinical candidates

Compute Requirements

Training cluster8× H100 SXM 80GB
Monthly GPU-hours12,000 GPU-hrs / mo
ProtLM training14,400 GPU-hrs / run
Diffuse training6,400 GPU-hrs / run
MolGNN training2,800 GPU-hrs / run
Inference (screening)~2,000 GPU-hrs / mo
Peak memory / GPU68GB (ProtLM, FSDP)
Batch size (ProtLM)64 / GPU · accum=4
InterconnectNVLink 4.0 + IB HDR
FrameworkPyTorch 2.4 · CUDA 12.5
Inference runtimeTensorRT · Triton
Screening throughput2.4M mols / 48h
05 / Benchmarks

Validated against standard
biomedical benchmarks.

We evaluate our models on established benchmarks across molecular property prediction (MoleculeNet), protein structure prediction (CAS-14), and molecular generation quality (GuacaMol). All numbers are from single-model evaluation without ensembling.

MoleculeNet (Avg AUROC)
89.2%
Average AUROC across 12 MoleculeNet tasks (Tox21, HIV, PCBA, MUV, BACE, BBBP, etc.). State-of-the-art for graph-based molecular property prediction without pretraining.
Protein Contact Prediction
0.83
F1 score on long-range contact prediction (CAS-14 targets). ProtLM learns structural contacts from sequence alone — no MSA or template information required.
Docking Success Rate
71.4%
Percentage of diffusion-generated molecules that achieve valid docking poses (RMSD < 2.0 Å) against held-out target binding pockets. Measured on 10K generated candidates.
Benchmark Metric Baseline Dynotx Delta
MoleculeNet (Tox21) AUROC 0.812 0.914 +12.6%
MoleculeNet (HIV) AUC-ROC 0.764 0.847 +10.9%
MoleculeNet (BACE) AUROC 0.858 0.921 +7.3%
MoleculeNet (BBBP) AUROC 0.729 0.868 +19.1%
Protein Contact (CAS-14) F1 (long-range) 0.72 0.83 +15.3%
GuacaMol (Validity) % valid molecules 0.870 0.966 +11.0%
GuacaMol (Novelty) % novel scaffolds 0.802 0.904 +12.7%
Docking Success % RMSD < 2.0 Å 52.8% 71.4% +35.2%
Inference Latency ms / molecule 340ms 47ms −86.2%
06 / About

Compressing drug discovery
from years to months.

Dynotx is an AI-driven biotechnology company using GPU-accelerated deep learning to discover and develop novel therapeutics. We build graph neural networks, protein language models, and diffusion-based generative models that screen millions of compounds, predict molecular properties, and generate novel drug candidates — in silico, at a scale that traditional drug discovery cannot match.

Traditional drug discovery takes 10–15 years and costs over $2 billion per approved drug. Our platform compresses the early discovery phase — target identification, hit screening, and lead optimization — from years to months by replacing wet-lab screening with GPU-accelerated virtual screening and generative molecular design.

We are a NVIDIA Inception member with access to NVIDIA's latest GPU architecture, cuDNN optimizations, and TensorRT inference tools. Our training infrastructure runs on H100 SXM clusters with FSDP distributed training, and we are actively scaling our compute capacity to support expanded screening campaigns and larger protein language models.

Three drug candidates from our pipeline are currently in preclinical evaluation: DTX-201 (oncology, KRAS G12C inhibitor), DTX-114 (antiviral, RNA-dependent RNA polymerase target), and DTX-307 (antibacterial, novel β-lactamase inhibitor scaffold).

contact@dynotx.tech

Founded
2024
Status
NVIDIA Inception Member
Focus
AI drug discovery · protein engineering · computational diagnostics
Pipeline
3 preclinical candidates (DTX-201, DTX-114, DTX-307)
Training Infrastructure
8× H100 SXM 80GB
NVLink 4.0 + InfiniBand HDR
FSDP · bf16 · PyTorch 2.4
Compute Scale
12,000 GPU-hrs / month
2.4M molecules screened
50K generated / target
Tech Stack
PyTorch · PyG · CUDA 12.5
TensorRT · Triton · RDKit
AlphaFold2 · OpenMM