Seeking Postdoctoral Positions · Jan 2027

Rajan Gyawali

PhD Candidate in Computer Science

Bioinformatics & Machine Learning Lab · University of Missouri–Columbia
Advisor: Dr. Jianlin Cheng

Google Scholar GitHub LinkedIn

rgkg2[at]missouri[dot]edu Columbia, MO

I am a computational biology and machine learning researcher developing deep learning methods at the intersection of AI and structural biology. My doctoral work centers on cryo-electron microscopy (cryo-EM) — designing neural architectures that automate protein particle picking, enable 3D atomic modeling by fusing cryo-EM data with AlphaFold3 predictions, and reduce annotation bottlenecks through few-shot and foundation model-based learning. My methods draw on computer vision, geometric deep learning, and multimodal learning, with applications spanning protein structure determination, virus characterization, and structure-based drug discovery.

My research is published in leading journals like Communications Chemistry, Scientific Data, Briefings in Bioinformatics and Bioinformatics, with benchmark dataset CryoPPP accumulating 55+ citations and adoption across the global cryo-EM community.

Research

MICA workflow — Multimodal integration of cryo-EM density maps & AlphaFold3 structural predictions

Project 01

3D Atomic Modeling · MICA

Multimodal Deep Learning for Protein Complex Structure Determination

MICA integrates cryo-EM density maps with AlphaFold3 structural predictions using a multi-task encoder-decoder with feature pyramid networks. It predicts backbone atoms, Cα atoms, and amino acid types to build an atomic model of protein complexes — achieving an average TM-score of 0.93 on high-resolution maps, outperforming all current state-of-the-art methods in both accuracy and model completeness.

Methods

Encoder-Decoder Feature Pyramid Network AlphaFold3 Multi-task Learning

Publication

MICA

Communications Chemistry (Nature), 2025

↗ Paper ↗ Code

CryoSegNet — Attention-gated U-Net + SAM particle picking pipeline

Project 02

Cryo-EM Particle Picking

AI-Driven Protein Particle Picking: CryoSegNet, CryoTransformer & CryoFSL

A suite of deep learning methods for automated cryo-EM particle picking. CryoSegNet integrates Meta's SAM with an attention-gated U-Net, achieving 3.32 Å average resolution — 14% better than crYOLO and 7% better than Topaz. CryoTransformer achieves an F1-score of 0.747 on diverse proteins. CryoFSL reduces annotation cost by 95% through few-shot learning, generalizing across protein types with minimal labeled data.

Methods

Segment Anything Attention U-Net Vision Transformers Few-Shot Learning

Publications

CryoSegNet

Briefings in Bioinformatics, 2024

↗ Paper ↗ Code

CryoTransformer

Bioinformatics, 2024

↗ Paper ↗ Code

CryoPPP — Dataset creation pipeline (Scientific Data, 2023)

Project 03

Benchmark Datasets

Community-Standard Cryo-EM Benchmarks: CryoPPP & CryoVirusDB

CryoPPP is a 2.6 TB benchmark of 9,893 expert-labeled cryo-EM micrographs with 300K+ particle annotations across 34 protein types, validated by 2D class analysis and 3D density map reconstruction. Community-standard dataset with 55+ citations. CryoVirusDB extends this to virus particles, providing the first annotated cryo-EM dataset for AI-based identification of icosahedral virus structures.

Methods

Data Curation Expert Annotation 3D Validation EMPIAR

Publications

CryoPPP

Scientific Data (Nature), 2023

↗ Paper ↗ Code

CryoVirusDB

Viruses, 2026

↗ Paper ↗ Code

Publications

CryoVirusDB: An Annotated Dataset for AI-Based Virus Particle Identification in Cryo-EM Micrographs

Gyawali R., Dhakal A., Wang L., Cheng J.

Viruses, 18(2), 224 · 2026

Journal Dataset ↗ Article ↗ Code

Multimodal Deep Learning Integration of Cryo-EM and AlphaFold3 for High-Accuracy Protein Structure Determination

Gyawali R., Dhakal A., Cheng J.

Communications Chemistry, 8(1) · 2025

Journal ↗ Article ↗ Code

Artificial Intelligence in Cryo-EM Protein Particle Picking: Recent Advances and Remaining Challenges

Dhakal A.*, Gyawali R.*, Wang L., Cheng J. (*equal contribution)

Briefings in Bioinformatics, 26(1) · 2025

Journal · Review ↗ Article

CryoFSL: An Annotation-Efficient, Few-Shot Learning Framework for Robust Protein Particle Picking in Cryo-EM Micrographs

Poudel B., Gyawali R., Dhakal A., Cheng J., Xu D.

bioRxiv · 2025

Preprint

CryoSegNet: Accurate Cryo-EM Protein Particle Picking by Integrating the Foundational AI Image Segmentation Model and Attention-Gated U-Net

Gyawali R., Dhakal A., Wang L., Cheng J.

Briefings in Bioinformatics, 25(4) · 2024

Journal ↗ Article ↗ Code

CryoTransformer: A Transformer Model for Picking Protein Particles from Cryo-EM Micrographs

Dhakal A., Gyawali R., Wang L., Cheng J.

Bioinformatics, 40(3) · 2024

Journal ↗ Article ↗ Code

Adapting Segment Anything Model (SAM) through Prompt-Based Learning for Enhanced Protein Identification in Cryo-EM Micrographs

He F., Yang Z., Gao M., Poudel B., Dhas N.S.E.S., Gyawali R., Dhakal A., et al.

IEEE International Conference on Medical Artificial Intelligence (MedAI) · 2024

Conference ↗ Paper

A Large Expert-Curated Cryo-EM Image Dataset for Machine Learning Protein Particle Picking

Dhakal A.*, Gyawali R.*, Wang L., Cheng J. (*equal contribution)

Scientific Data (Nature), 10(1) · 2023

Journal Dataset · 55+ citations ↗ Article ↗ Code

Predicting Protein-Ligand Binding Structure Using E(n) Equivariant Graph Neural Networks

Dhakal A.*, Gyawali R.*, Cheng J. (*equal contribution)

bioRxiv · 2023

Preprint

An Approach for the Employee Face Recognition by RPN and Faster R-CNN Techniques

Gyawali R., Pant D.R.

Proceedings of IOE Graduate Conference, Vol. 6 · 2019

Conference ↗ Paper

Experience

Aug 2022 – Present

Graduate Researcher

University of Missouri — Bioinformatics & Machine Learning Lab

Columbia, MO

Developed MICA, a multimodal deep learning method integrating cryo-EM density maps and AlphaFold3 for automated 3D atomic modeling via encoder-decoder + feature pyramid networks, outperforming all state-of-the-art methods.
Designed CryoSegNet, an attention-gated U-Net + SAM framework achieving 7–14% improvement in 3D reconstruction resolution (3.32 Å) over crYOLO and Topaz.
Curated CryoPPP, a 2.6 TB benchmark (9,893 micrographs, 300K+ labels, 34 proteins) — the community-standard cryo-EM dataset with 55+ citations.
Curated CryoVirusDB, an annotated cryo-EM dataset for AI-based icosahedral virus particle identification and classification.
Contributed to CryoFSL, CryoTransformer, and prompt-based SAM learning for cryo-EM protein identification.
Trained GNNs, diffusion models, and detection transformers using multi-node multi-GPU DDP on HPC clusters (Slurm, LSF); built large-scale parallel data preprocessing pipelines.

May 2018 – Jun 2022

Telecommunications Engineer

Nepal Telecom

Kathmandu, Nepal

Built a full-stack Django/REST KPI analytics tool for automated analysis across 5,000+ network elements, reducing analysis time from days to minutes.
Designed Python/SQL data pipelines processing 10+ GB daily of network performance data with statistical analysis and real-time visualization.
Developed LSTM-based predictive models for network anomaly detection and performance forecasting, improving reliability by 15%.
Automated reporting workflows reducing manual workload by 80%.

Oct 2016 – Jun 2022

Lecturer & Adjunct Lecturer

Himalaya College of Engineering, Tribhuvan University

Kathmandu, Nepal

Taught undergraduate courses to classes of up to 48 students: C Programming, OOP, Computer Graphics, Image Processing, Computer Networks, Signal Processing, and Communication Systems.
Delivered Python Programming and Data Analysis & Visualization training to Computer and Electronics Engineering undergraduates.
Supervised final year projects; led weekly lab and discussion sections for groups of 16–24 students.

Education

Aug 2022 – Dec 2026 (Expected)

GPA 4.0

PhD, Computer Science

University of Missouri–Columbia, USA

Emphasis: Deep Learning, Computational Biology & Bioinformatics
Advisor: Dr. Jianlin Cheng
Dean's Fellowship recipient, 2022

2017 – 2019

Score: 90.13%

MSc, Information and Communication Engineering

Tribhuvan University, Nepal

Thesis: Employee Face Recognition by RPN and Faster R-CNN — 96.0% accuracy on Chokepoint dataset
Project: CNN for Multiclass Face Recognition — 96.44% accuracy, 13K images, 5,749 identities
Graduate Scholarship recipient, 2017

2012 – 2016

Score: 82.06%

BE, Electronics and Communication Engineering

Tribhuvan University, Nepal

Final Project: Brain Controlled Wheelchair using BCI & NeuroSky Mindwave
Undergraduate Scholarship recipient, 2012

Skills

Deep Learning

CNNsVision TransformersU-NetResNetAttention MechanismsEncoder-DecoderGNNsE(n)-Equivariant NetworksDiffusion ModelsDetection TransformersFew-Shot LearningPrompt-Based LearningTransfer Learning

Frameworks

PyTorchPyTorch LightningPyTorch GeometricDGLTensorFlowscikit-learnNumPyPandasSciPyMatplotlibPlotly

Structural Biology

Cryo-EMProtein Structure PredictionProtein-Ligand BindingAlphaFoldRELIONCryoSPARCUCSF ChimeraXPyMOLPyRosettaBiopythonRDKit

HPC & Distributed

Multi-GPU DDPMulti-Node TrainingSlurmLSFParallel Processing

Languages

Python (Expert)C++BashRSQL

Dev Tools

Git / GitHubDockerLinuxWeights & BiasesTensorBoardJupyterDjango

Talks & Presentations

2024

CryoSegNet: AI-based method for protein particle picking from cryo-EM density micrographs

Cryo-EM Super Group · University of Missouri, Columbia, MO

2024

A large expert-curated cryo-EM image dataset for machine learning protein particle picking

RECOMB Conference (Poster Presentation) · Cambridge, MA

Awards & Service

Honors & Fellowships

2022Dean's Fellowship — University of Missouri
2017Graduate Scholarship — Tribhuvan University
2012Undergraduate Scholarship — Tribhuvan University

Service

Manuscript Reviewer — NeurIPS, Briefings in Bioinformatics, BMC Bioinformatics, Communications Chemistry, etc.
President — University of Missouri Nepalese Students Association (May 2025 – Apr 2026)
Treasurer — University of Missouri Nepalese Students Association (May 2024 – Apr 2025)
Member — Academic Affairs Committee, University of Missouri (Aug 2023 – May 2024)