School of Biomedical Sciences cordially invites you to join the following Post-doctoral Fellow (PDF) Seminar:

Date: 17 June 2025 (Tuesday)
Time: 4:00 pm – 5:00 pm
Venue: Seminar Room 4, G/F, Laboratory Block, 21 Sassoon Road
Host: Dr. Haifeng Fu

Light refreshments will be served. Please register via the below link by 16 June 2025 (Monday):
Registration: https://hku.au1.qualtrics.com/jfe/form/SV_8qBBzKYwxLfY2Bo

Characterization of transcriptional regulations of telomerase in cancer and stem cells
Dr. Lap Hang Tse (Post-doctoral Fellow)
[Supervisor: Professor David Shih]

Telomerase reverse transcriptase (TERT) play crucial roles in maintaining chromosomal stability through telomere elongation, conferring cells the capacity for unlimited replication. While TERT expression is predominantly restricted to stem cells and absent in differentiated cells due to transcriptional repression, its re-activation representing a hallmark of oncogenic transformation. Elevated expression of TERT in cancer is frequently associated with promoter mutations or methylations that affect the transcription regulation. We aim to develop statistical models to characterize the transcriptional regulations of TERT across different cell types. By identifying the differences in TERT transcriptional regulation between cancer and stem cells, we hope to provide insights for potential therapeutic strategies targeting TERT dysregulation in cancer cells while preserving its physiological function in normal stem cells.

Machine-learning guided low-N search for top variants for genome editor engineering
Dr. Hoi Yee Athena Chu (Post-doctoral Fellow)
[Supervisor: Professor Alan Wong]

We developed a strategy to obtain the greatest number of best-performing protein variants with the least amount of experimental effort for mutagenesis screens to alleviate the experimental resources spent on cloning and testing non-functional variants. Our strategy uses zero-shot prediction and machine learning to guide multi-round sampling of top variants in the library. We found that four rounds of low-N pick-and-validate sampling with 12 variants for machine learning yielded the up to 92.6% accuracy in selecting the true top 1% variants mutant libraries, while two rounds of 24 variants identify top variants with higher sequence diversity. Our proposed strategy outperforms other state-of-the-art methods in terms of efficiency and accuracy and can be generalized for a range of protein-function inference including Cas9 and APOBEC protein in the CRISPR genome editor toolkit.

All are welcome.