Deep Learning for Sequences & Structures — Hands-on

Design end-to-end deep learning workflows for biological sequences and macromolecular structures. You will implement modern architectures (CNNs, Transformers, GNNs) , leverage protein language models, apply self-supervised objectives, and deliver reproducible experiments with clear interpretability.

Deep Learning for Sequences & Structures

Help Desk · WhatsApp

Session Index

Session 1 — DL Foundations for Bio Sequences Session 2 — Transformers & Protein Language Models Session 3 — Graph & Structural DL (3D/Contacts) Session 4 — Mini Capstone: Fine-tune & Explain

Session 1

Fee: Rs 10,800

DL Foundations for Bio Sequences

Data representations & batching

one-hot/k-mer/embeddings tokenization class imbalance

Architectures & training

CNN / dilated convs RNN/GRU/LSTM optimizers/schedules

Toolchain

PyTorch / Lightning TensorFlow/Keras Weights & Biases

Session 2

Fee: Rs 14,600

Transformers & Protein Language Models

Attention & transformer stacks for bio

positional encodings masked LM sequence-to-function

Protein LMs & fine-tuning

ESM / ProtBERT TAPE / ProGen adapters/LoRA

Efficiency & scaling

mixed precision gradient checkpointing distributed training

Session 3

Fee: Rs 18,800

Graph & Structural DL (3D/Contacts)

Graphs & 3D representations

contact/distance maps equivariant nets (SE (3) ) message passing

GNN toolkits & tasks

PyTorch Geometric DGL structure property prediction

Regularization & robustness

dropout/weight decay augmentation adversarial tests

Session 4

Fee: Rs 23,600

Mini Capstone: Fine-tune & Explain

Fine-tune a protein LM or GNN for a real task

Theory + Practical

Interpretability & reporting

saliency/CAM/integrated gradients token/edge attributions model cards

Deliverables: code, metrics dashboard & brief

notebook/script ROC/PR plots PDF/HTML

PDF

Deep Learning for Sequences & Structures Training | Transformers, Protein LMs, GNNs, Interpretability

Deep Learning for Sequences & Structures — Hands-on