01
What is RNA?
RNA (ribonucleic acid) is a biological molecule made from nucleotides. Its sequence and ability to fold into specific structures are central to many biological processes.
Computational Structural Biology
Understanding RNA
through computation.
UMFold is a research algorithm for RNA tertiary structure prediction, developed with a focus on computational optimization and parallel computing.
THE RESEARCH
From a structural challenge
From a structural challenge
to a computational approach.
RNA function is closely connected to molecular structure. UMFold investigates a computationally optimized approach for predicting RNA tertiary structures, with particular attention to longer sequences and execution time.
UNDERSTANDING RNA STRUCTURE
From sequence to three-dimensional structure
RNA can be understood at several structural levels. The sequence provides the molecular information, secondary structure describes important base-pairing patterns, and tertiary structure represents the final three-dimensional folding of the molecule.
02
Primary Sequence
The primary structure is the linear nucleotide sequence of an RNA molecule. It records the order of its bases—adenine (A), uracil (U), guanine (G) and cytosine (C).
03
Secondary Structure
The secondary structure describes how nucleotides pair and form structural regions such as hairpins, loops, bulges and stems. UMFold uses the corresponding dot-bracket secondary-structure information as part of its input.
Why is it important?It provides the structural framework from which loops, free bases and potential tertiary interactions can be identified.
04
Tertiary Structure
The tertiary structure is the complete three-dimensional arrangement produced when the RNA folds in space, including interactions between structural regions. UMFold predicts this 3D structure in PDB format.
Why is it important?RNA functions in many biological applications are dependent on its three-dimensional structure, making tertiary-structure prediction an important computational problem.
Primary sequence→Secondary structure→UMFold computational modelling→Tertiary 3D structure
RESEARCHER
Bioinformatics · HPC
Bioinformatics · HPC
THE RESEARCHERView researcher profile →
Mrs. Ujjwala Hemant Mandekar
Academic researcher working across bioinformatics, computational optimization, parallel algorithms and high-performance computing.
B.E.Computer Technology · 1999
M.Tech.Computer Science & Engineering · 2009
Ph.D.Computer Science & Technology · 2021
ResearchRNA tertiary structure prediction
CurrentLecturer, Government Polytechnic Sakoli
WHY THIS RESEARCH?
Understanding the challenge behind RNA tertiary structure prediction.
RNA structure is closely connected to its biological function. Predicting the tertiary structure computationally becomes particularly challenging as RNA sequences become longer and the structural search becomes more computationally demanding.
Long RNA sequences
The research identifies a practical need to work with RNA sequences much longer than the limits reported for several existing prediction approaches.
Computational cost
RNA tertiary-structure prediction can require substantial computation. The research therefore focuses on reducing execution time and using computational resources efficiently.
The research objective
UMFold was developed as a computationally optimized approach for tertiary-structure prediction, followed by parallel execution on shared-memory multicore architectures.
RESEARCH
The UMFold approach
UMFold takes primary sequence and corresponding secondary-structure information and constructs a predicted RNA tertiary structure in PDB format.
01RNA inputPrimary sequence + dot-bracket secondary structure
→
02Structural analysisLoops, base pairs and free bases
→
03Tertiary interactionsPseudoknots, kissing hairpins, hairpin–bulge
→
04PDB searchSimilarity matching and best configuration
→
053D structurePredicted tertiary structure in PDB
Structural regions
Hairpin, internal, bulge, stack, single-stranded and multi-loop regions are considered during structural analysis.
Similarity-driven modelling
Candidate structural regions are searched against PDB metadata and the configuration with the highest similarity is selected.
Computational optimization
The tertiary interaction procedure uses dynamic programming with a slight divide-and-conquer approach, followed by parallelization.
PARALLEL COMPUTING
Reducing computation time with OpenMP.
Parallel computing is a central part of the UMFold research. The algorithm was implemented on shared-memory multicore architectures to accelerate computationally intensive stages of RNA tertiary-structure prediction.
Why parallelize UMFold?
After tertiary interactions are identified, candidate loop subsequences must be searched against approximately 2,753 PDB Meta files. The research identifies this search and the subsequent construction of the tertiary PDB file as the most time-consuming procedures.
Where OpenMP is applied
01 · PDB Meta search
Loop subsequences are distributed across available processors to find matching structural regions and their similarity values.
02 · Tertiary PDB construction
Matching PDB data associated with different PDB IDs is processed in parallel to construct the resultant tertiary structure.
03 · Memory-aware processing
A UniquePDBID vector helps avoid repeatedly loading the same PDB file when multiple loop-similarity results refer to it.
04 · Dynamic scheduling
The documented implementation uses OpenMP dynamic scheduling with a chunk size of 100 for the PDB Meta-file search.
Shared-memory execution
The sequential workflow is retained while selected computationally expensive portions are parallelized. OpenMP creates worker threads, distributes suitable loop iterations among processors, and joins the results before the next stage of the algorithm.
REPORTED SPEEDUP
4 cores≈ 1.90×
8 cores≈ 2.63×
12 cores≈ 3.66×
28 cores≈ 7.37×
For experiments on sequences extending to approximately 10,000 nt, the thesis reports about 7.37× speedup on a 28-core machine.
APPLICATIONS
Where this research can be useful
The applications below are grounded in the scope and experiments described in the research—not claims of clinical or commercial deployment.
Long RNA structure prediction
Computational prediction for longer RNA sequences, with experiments reported up to approximately 10,000 nucleotides.
Viral RNA analysis
Computational analysis of long RNA sequences relevant to viral RNA research.
RNA bioinformatics
Connecting secondary-structure information with computationally predicted tertiary structures.
Tertiary interaction analysis
Analysis of structural regions and possible interactions such as pseudoknots and kissing hairpins.
High-performance RNA computing
Demonstrating shared-memory parallel techniques for computationally intensive RNA analysis.
Computational structural biology
Generation and evaluation of RNA 3D structures using structural databases, similarity and RMSD analysis.
EXPERIMENTAL RESULTS
Evidence from the research
Reported experimental results demonstrate both structural evaluation and computational performance.
7.37×Speedup on 28 cores
≈10,000nt sequence length reported
2.6391Average UMFold RMSD
100%Reported comparison-set success rate
Speedup by core count
Reported valuesAverage RMSD comparison
Lower is betterUMFold2.63910714
RNA Composer2.63946429
iFoldRNA2.65269310
VFold2.65882716
The thesis reports RMSD values below the stated permissible value of 6.
PUBLICATIONS
Google Scholar ↗Scholarly contributions
Research publications and academic contributions associated with Mrs. Ujjwala Hemant Mandekar.
An Evolutionary Algorithm based solution for Register Allocation for Embedded Systems
Ujjwala H. Mandekar
Explores an evolutionary-algorithm-based approach to register allocation for embedded systems, focusing on efficient assignment of processor registers during compilation.
Research paperPrediction of RNA 3D Structure
Ujjwala H. Mandekar · S. P. Khandait
Presents computational work on predicting the three-dimensional structure of RNA, contributing to the broader problem of RNA structural modelling.
Journal paper · RNA / BioinformaticsPredicting RNA Tertiary Structure Using Parallel Algorithm
Ujjwala H. Mandekar · Leena Patil · Sunanda Khandait
Describes an approach for RNA tertiary-structure prediction that uses parallel computation to improve the execution of a computationally intensive structural-analysis task.
Journal paper · RNA / Parallel ComputingComputational Optimization and Analysis of Functional Bioinformatics using Parallel Algorithm
Ujjwala Hemant Mandekar · Sunanda P. Khandait · Leena H. Patil
Focuses on computational optimization and parallel algorithms for functional bioinformatics, closely aligned with the research direction behind UMFold.
Journal paper · Bioinformatics / HPCumFold: An Algorithm to Predict RNA Tertiary
Ujjwala H. Mandekar · Sunanda Prabhakar Khandait
Presents UMFold as an algorithm for RNA tertiary-structure prediction, describing the computational approach developed for structural modelling.
Book / research chapterIntegration of Machine Learning and Process Analytical Technologies (PAT) in Food Industry
Vijaya P. Balpande · Ujjwala Hemant Mandekar · Pravin B. Pokle · Ajay M. Mendhe · M. G. Pathan
Examines the integration of machine learning with Process Analytical Technologies (PAT) in the food industry, connecting data-driven methods with process monitoring and analysis.
Journal paper · AI / Food TechnologyEnhancing MQTT Security in the Internet of Things with an Enhanced Symmetric Algorithm
Rupali Atul Mahajan · Rupesh G. Mahajan · Manjusha Tatiya · Ujjwala Hemant Mandekar · Minal Shahakar · Yogendra Patil
Addresses MQTT security in Internet of Things environments and proposes an enhanced symmetric-algorithm approach for improving message-security protection.
Journal paper · IoT / SecurityIntroduction to High-Performance Computing Architectures
Ujjwala Mandekar · Vijaya P. Balpande · Pradnya Borkar · Pooja B. Aher
Introduces the architectural concepts underlying high-performance computing systems and provides an academic overview of HPC architectures.
Book chapter · HPCHigh-Performance Computing: Use Cases, APIs and Applications
Pradnya Borkar · Sagarkumar Badhiye · Ujjwala Mandekar · Vijaya P. Balpande · Roshani Raut · Pratik Agrawal
Covers high-performance computing use cases, APIs and application-oriented aspects of developing solutions for HPC environments.
Book chapter · HPCParallelization Techniques
Vijaya P. Balpande · Ujjwala Mandekar · Pradnya Borkar
Introduces techniques used to parallelize computational workloads, complementing the researcher's broader work in parallel and high-performance computing.
Book chapter · Parallel ComputingWeather sentiment driven quantum graph networks for intraday stock prediction
Pradnya Borkar · Vijaya Parag Balpande · Ujjwala Hemant Mandekar · Sthitipragyan Biswal · M. Karthikeyan
Explores the use of weather sentiment and quantum graph networks for intraday stock-prediction modelling, combining alternative data with advanced computational learning methods.
Research article · AI / Quantum Graph NetworksPublication metadata should be treated as a curated website list; the Google Scholar profile remains the authoritative external source for the live publication record.
RESEARCHER PROFILE
Teaching, research and computational innovation.
Mrs. Ujjwala Hemant Mandekar is an academic and researcher whose work connects computer science with bioinformatics, computational optimization and parallel computing.
Mrs. Ujjwala Hemant Mandekar
Lecturer, Government Polytechnic Sakoli, Maharashtra, India
Her research journey includes the development of UMFold, a computational approach for RNA tertiary structure prediction with emphasis on computational optimization, structural similarity and execution on shared-memory multicore architectures.
EDUCATION
Academic qualifications
B.E. — Computer Technology
Chandrapur Engineering College, R.T.M. Nagpur University · 1999
M.Tech. — Computer Science & Engineering
G.H. Raisoni College of Engineering, R.T.M. Nagpur University · 2009
Ph.D. — Computer Science & Technology
R.T.M. Nagpur University · 2021
DOCTORAL RESEARCH
Computational optimization in bioinformatics
Ph.D. — Computer Science & Technology
R.T.M. Nagpur University · 2021
Doctoral research at Priyadarshini Institute of Engineering and Technology on “Computational optimization and analysis of functional bioinformatics structures using parallel algorithm”, under the guidance of Dr. Sunanda P. Khandait.
The completed doctoral research led to the development of a method for RNA tertiary structure prediction and the UMFold research work presented on this website.
RESEARCH INTERESTS
Areas of academic interest
- RNA tertiary structure prediction
- Bioinformatics and structural analysis
- Computational optimization
- Parallel algorithms and OpenMP
- High-performance computing
- Computational structural biology
ACADEMIC & PROFESSIONAL JOURNEY
From computer technology education to computational bioinformatics research.
2000
Lecturer
Started academic career at Priyadarshini Polytechnic, Nagpur.
2006
Head, Computer Technology Department
Promoted to departmental leadership at Priyadarshini Polytechnic, Nagpur.
2010
Assistant Professor
Joined Priyadarshini J.L. College of Engineering, Nagpur.
2016
Government service
Joined a government organization through MPSC selection.
Present
Lecturer
Government Polytechnic Sakoli, Maharashtra.
CONTACT
Open Google Scholar ↗Explore the research.
For academic enquiries, collaborations or further information about the research, please use the appropriate institutional contact channel.