National Agricultural Innovation Project (NAIP), ICAR and the International Food Policy Research Institute (IFPRI) organized a two day workshop on ‘Impact of capacity building programs under NAIP’ on June 6-7, 2014 at AP Shinde Auditorium, NASC Complex, Pusa, New Delhi. The main purpose of the workshop was to present and discuss the findings of the impact evaluation study on capacity building programs under NAIP by IFPRI. The scientists from ICAR and agricultural universities were sent abroad to receive training in specialized research techniques. Post-training, scientists were expected to work on collaborative projects within the ICAR, which would further enrich their knowledge and skills, expand their research network and stimulate them’ to improve their productivity, creativity and quality of their research. The ICAR commissioned with IFPRI (International Food Policy Research Institute) to undertake an evaluation of these capacity building programs under NAIP in July 2012. The workshop shared the findings on the impact of capacity building programs under NAIP and evolve strategies for future capacity building programs
IFPRI - NAIP - Reflections of Researchers on Capacity Building Programs - A R Rao
1. Reflections of Researchers on Capacity Building Programs
International Open Training
A. R. Rao
Centre for Agricultural Bioinformatics
Indian Agricultural Statistics Research Institute,
New Delhi
Workshop on Impact of Capacity Building Programs under NAIP
NASC, Pusa, New Delhi
June 6-7, 2014
2. General Information
• Name : A.R.Rao
• Organization : IASRI
• Duration : 24-02-11 to 23-05-11
• Area of training : Bioinformatics
• Place of training : Univ. of Washington, Seattle
• Expenditure : Rs.5,01,836
3. Skill Gaps
• Institutional: bridge the gap between genomic information
and knowledge by utilizing statistical and computational
sciences (Bioinformatics)
• Individual: Infer knowledge from quantitative analysis of
genome sequence data by application of modern and
algorithmic approaches
4. Capacity building, Work Culture & Attitudinal Changes
• Genome Prediction and its application in (i) Disease risk
prediction (ii) prediction of breeding values or trait performance
from SNP genotyping data
• Methodology: Least Absolute Shrinkage Selection Operator
(LASSO) – Penalized Regression Method; K-fold cross validation
• Advanced methods – Machine Learning Approaches: Random
Forest
5. Capacity building, Work Culture & Attitudinal Changes
• Share resources and data (avoid undertakings)
• Interact more with biologists and application scientists
• Listen to webinars and understand language of others
• Give importance to others’ time
• Increase transparency in work
• Constant monitoring of self and other staff engaged in
projects
6. Capacity building, Work Culture & Attitudinal Changes
• listen carefully, ask to repeat if not followed
• think and talk precise, speak with clarity and facts
• Commit only to the extent which can be delivered in time
• Plan before start of work
• Do not hesitate to ask questions like “why? and what?”
7. Planning, Selection & Training duration
• Well planned by the funding agency in terms of
announcement, format of application and timely processing
of proposals
• Highly satisfied with the selection process
• For effective training – duration would have been 6 months
(duration depends on the requirement of the host institute,
sponsoring institute and interest of the candidate)
8. Resource Institution and Resource Person
• University of Washington, Seattle
• Bruce Weir, Professor & Chair, Biostatistics
Department, Adjunct Professor, Genome
Sciences, University of Washington, Seattle, USA
• Research Interests: Statistical methodology for
genetic data, population structure, disease
associations and relationships, use of genetic
data for human identification
9. Training Implementation and Reporting
• Introduction and topic assignment
• Review of literature and understanding of topic
• Understanding UW’s High Performance Computing
Environment
• Getting access to the system, raw data and refined data
• Understanding data kept in Network Common Data Form
(NetCDF) – Reading through notes and manuals – discussion
through emails – rare in-person discussions
10. Training Implementation and Reporting
• Samples genotyped on the Illumina 1M beadchip
• Understanding of methodology
• Writing perl and R scripts for parsing and analysis of data,
debugging and final execution of programmes
• Submission of jobs in queue using HPC cluster
• Processing of results and drawing inference
• Discussion with fellow colleagues
• Attending seminars, webinars, meetings; Participation in audio
and video conferencing
• Reading training manuals of Summer Institutes; Report Writing
11. Post-Training Utilization
Research
• Developed a new approach for mini-core identification and
illustrated it by using data generated under Bioprospecting of
genes and allele mining for abiotic stress tolerance project
• Compared the performance of LASSO with other methods
available for classification and prediction purposes
• Guided 1 Ph.D. and 1 M.Sc. Students in the trained area
• Tried on SNP-genotyping data of maize generated under BAM
project
• Applied on data, where the number of variables (p) are larger
than number of observations (n), particularly in case-control
studies
12. Post-Training Utilization
Training
• Recent advances in statistical and computational genomics
data analysis (19th - 28th March, 2012)
• ICAR-Winter School on Recent advances in quantitative
genetics and statistical genomics during (06-26th November,
2012)
• Advanced Analytical Techniques in Bioinformatics (10-19th
March, 2014)
• Several lectures were delivered on Genome Prediction in
various trainings organized under NARS.
13. Future Plan
• Interactions will be held with molecular biologists, plant and
animal breeders interested in Genomic Prediction. To play a
significant role in the upcoming network platforms.
• Application of penalized regression methods for prediction of
economically important traits like milk yield, body mass index,
lactation length, calving interval, etc.
• Application of Lasso and elasticnet methods in disease risk
prediction in plant, animal and fish species.
• Application of Random Forest methodology and algorithmic
based machine learning approaches for genome prediction
• High Performance Cluster management in National
Agricultural Bioinformatics Grid
14. Acknowledgements
• Director, IASRI
• ICAR, NAIP, World Bank, ND, NCs and Staff at NAIP Office
• Prof. Bruce S. Weir, Prof. & Chair, Biostatistics Department,
University of Washington, Seattle
• David Levein, Stephine Groton, Sarah Nelson, David Crosslin,
Rui Zhang, Rohit Swarnkar at GCC
• Director, NAARM