I have been asked to give a talk at the Workshop on Automated Algorithm Design being held at the Toyota Technical Institute at Chicago from August 7th to the 9th. I will be speaking about our work related to applying parameter advising to reference-based transcript assembly (see Publications).
While in Chicago, I will also attend the Workshop on Learning-Based Algorithms also being held at TTIC from August 12th to the 14th.
Category: Talks
Universal k-mer sets work accepted to ACM-BCB
Our recent work on universal k-mer sets has been accepted as a paper to The 10th ACM Conference on Bioinformatics, Computational Biology, and Health Informatics (ACM-BCB) in September, being held in Niagara Falls, NY. The work titled “Practical universal k-mer sets for minimizer schemes” is a collaboration with our summer iBRIC student from 2017 Fiyin Gbosibo, Carl Kingsford, and Guillaume Marçais. The preprint is now available (see Publications).
Toward building an automated bioinformatician: parameter advising for improved scientific discovery
Modern scientific software has a large number of tunable parameters that need to be adjusted to ensure computational performance and accuracy of the results. When these parameter choices are made incorrectly we may overlook significant results or falsely report insignificant ones. Optimizing the parameter choices for one input may not provide an assignment that’s good for another, so this parameter optimization process typically needs to be repeated for each new piece of data. Standard machine learning methods for solving this problem need to repeatedly run the software which may not be suitable in practice. Because of the time consumption required to optimize parameters and the possible loss of accuracy that can result when chosen incorrectly, the default parameter vector that are provided by the tool developer is often used. These defaults are designed to work well on average, but most interesting cases are rarely “average”.
In this talk, I describe my first steps in automatically learning the correct program configuration for biological applications using a framework we call “Parameter Advising”. To apply this framework to the problem of multiple sequence alignment we developed an accuracy estimator, called Facet, to help choose alignments since no ground truth is available in practice. When we use Facet for advising on the Opal aligner we boost accuracy by 14.6% on the hardest-to-align benchmarks. For the reference-based transcript assembly problem, when applying parameter advising to the Scallop assembler we see an increase in accuracy of 28.9%. The framework is general and can be extended to other problems in computational biology and beyond. I will discuss possible areas where parameter advising could be used to automatically learn to run complex analysis software.
Parametric sequence alignment
This lecture describes parametric sequence alignment as presented in Section 13.1 of Dan Gusfield’s book Algorithms on Strings, Trees, and Sequences.
Given two sequences, can you determine how many choices of values of the standard sequence alignment objective function’s parameters give distinct optimal alignments. In other words, how many alignments are optimal for some setting of the objective function’s parameters. This lecture discusses how to approach and answer this question.
Biological Data Science at Cold Spring Harbor Laboratory
I have been invited to give a talk at the Cold Spring Harbor Laboratory Biological Data Science meeting (#biodata18) November 7-10. My talk is preliminarily titled “Building an automated bioinformatician—More accurate, large-scale genomic discovery using parameter advising”.
Continue reading “Biological Data Science at Cold Spring Harbor Laboratory”