Women in computing: Yuxuan Lan
Working at the intersection of technology and biology, Dr Yuxuan Lan is bringing her wealth of experience in machine learning to the analysis of complex genomes - developing real time analysis of long reads from the Oxford Nanopore MinION.
It was around this time that computing was really coming to the fore - not just in how we communicate, but in how we answer a swathe of scientific questions.
Hi Yuxuan, what are you currently working on?
I am working on a BBSRC funded project to create a general-purpose user friendly tool for real-time analysis of nanopore metagenomic data. Nanopore technology can generate long read sequences and allows real-time access during sequencing. Current software tools for real-time analysis are either limited in functionality or are not open-source or free.
The Leggett group is working to fill this gap. The goal is to make an open-source tool that has a plug-in interface for third party tools, is easy for visualisation, can be run on both HPC and low power computing equipment such as a laptop or computer stick to facilitate in-field and point of care real-time analysis, and so on.
How did you get into computing?
I studied electronics and telecommunications for my undergrad degree. I did a dissertation on designing an office information system which I really enjoyed and decided to learn more about computing.
Yuxuan started her career in electronics and telecommunications before switching to bioinformatics.
What sort of projects have you worked on?
Before coming to bioinformatics, I was working on computer lip-reading, i.e., automatically recognising speech with only visual input. Imagine trying to buy a train ticket from a kiosk. Instead of having to go through many steps on the screen, wouldn’t it be nice to just tell the computer through a microphone where you want to go?
However, the station may be noisy and this could make an audio speech recognition system ineffective.
Equipped with also a webcam, a computer can understand you better by reading your lips when the surroundings are very noisy, or by both listening to and seeing what you say when it is quieter.
Before coming to bioinformatics, I was working on computer lip-reading - automatically recognising speech with only visual input.
How has machine learning progressed since you did your PhD?
There are many advances - and a significant one is the deep learning. I started my PhD before deep learning started to develop.
At the time, learning algorithms were important but a lot of focus was on designing and selecting features for the algorithm. Now the feature selection can be done by the deep learning.
Who wins at lip reading - robots or people?
There is no clear winner. Only expert humans can lip-read, but very good lip-readers can outperform computers. We found that many lip-readers heavily rely on the language model, i.e., context, in order to lip-read. This makes them perform well on a familiar topic but struggle on an alien one.
Computers on the other hand don’t have such concerns - they can train on many topics. Their performance falls between the best and the worst expert lip-readers, based on a paper we published. I believe more progress has been made in the last few years, with the continuous development in machine learning, especially in the area of deep learning.
Computers on the other hand don’t have such concerns - they can train on many topics. Their performance falls between the best and the worst expert lip-readers...
What’s it like making the switch to bioinformatics?
It is a completely new field to me. It took me a few months to start to follow seminars and talks. It is also a fast moving field with people pushing the frontier of science all the time.
I feel it challenging but also enjoyable to be near the forefront of science and discovery.
How has it been balancing a young family with working in science and computing? Has the flexible working here at EI helped?
Ha, it has never been easy. My husband is very supportive and understanding, and more than happy to share the burden of childcare.
It helps that he and I both work at the Earlham Institute where, when it is possible, staff are offered the convenience of flexible working, and it helps us to work around the childcare, for example, taking my child to after school activity classes, or attending a sick child at home.
How’s your Chinese tea tasting social going? What’s the crowd favourite so far?!
I’ve done two tea tasting sessions so far. The first time we tried unsmoked LapSang Souchong, and the second time Dian Hong.
Both red teas were well received by the tasters, so I think red tea is the Brits’ favourite, as we all know.
Yuxuan runs a very informative Chinese tea tasting social club each month at EI.