@f_hach's lab. Based in Vancouver Prostate Centre. Computational algorithms for cancer genomics research. 🧬💻
Vancouver, British Columbia
Joined January 2019
- Tweets78
- Following52
- Followers146
- Likes35
TKSM is now accepted by Bioinformatics and published as an advance article!
doi.org/10.1093/bioinformati…
Take a moment to check this quite cool and very modular long-read simulator built by our PhD students @fkaraoglan_cs and @baraaorabi 👇🏻
Our latest manuscript that's spearheaded by @fkaraoglan_cs and @baraaorabi (with the co-supervision of Cedric Chauve and collaboration of @RyanFlannigan00) is out!
We are delighted for TKSM to say Hello World! 👋
💻: github.com/vpc-ccg/tksm
📄: biorxiv.org/content/10.1101/…
Hach Lab retweeted
Nice work from two @BilkentCompGen alumni!
Our latest manuscript that's spearheaded by @fkaraoglan_cs and @baraaorabi (with the co-supervision of Cedric Chauve and collaboration of @RyanFlannigan00) is out!
We are delighted for TKSM to say Hello World! 👋
💻: github.com/vpc-ccg/tksm
📄: biorxiv.org/content/10.1101/…
And of course, this work would not have been possible without the funding and institutional support of
@SFU_CompSci
@UBC_CS
@VanProstateCtr
@UBCUrology
@SFUMath
@NSERC_CRSNG
@HlthResearchBC
All of this is packaged into a neat package that comes with a handy Snakefile to run your simulation configs (including full UNIX piping support!).
And the insulation of each module means that you can write your own TKSM module! Here is a list of currently implemented modules:
These simulation pipelines can be more complex, with branches defined as their own linear pipelines! For example here is a scRNA-seq experiment that is made possbile by TSKM config:
This design allows the user to define arbitrary simulation pipelines using easy-to-write YAML configs. For example, here is a simple bulk RNA-seq long-read pipeline and its corresponding YAML config:
TKSM is a modular, easy-to-customize, and fast long-read simulator. Our design philosophy in TKSM is:
1. Each library prep step will be targeted by a TKSM module
2. The I/O of each core module is a description of nuclueic molecules that (eventually) will be reads
Our latest manuscript that's spearheaded by @fkaraoglan_cs and @baraaorabi (with the co-supervision of Cedric Chauve and collaboration of @RyanFlannigan00) is out!
We are delighted for TKSM to say Hello World! 👋
💻: github.com/vpc-ccg/tksm
📄: biorxiv.org/content/10.1101/…
Hach Lab retweeted
I am extremely proud to share that our recent method's manuscript on functional testing of non-coding genetic variants got published. doi.org/10.1093/nar/gkac1198. (1/6)
Congratulations on all your hard work and well deserved degrees! Can't wait to see what you go on to achieve next 🎉🥂
(when you have four grads defending in a row [including three in the same week!] mid-day celebration is absolutely a must!)
This semester has been filled with proud moments; we had the pleasure of FOUR grad students defending their theses!
MSc Ghazal Ebrahimi, MSc @ArmitaSafa, MSc Reza Soltani, and PhD @tuncmorova successfully defended their theses in @ubcscience Graduate Program in Bioinformatics!
This work was spearheaded by our PhD candidate @baraaorabi.
Thanks for all the support from @VanProstateCtr, @UBC_CS, @SFUMath @UBCUrology and the funding from @NSERC_CRSNG, @HlthResearchBC, @CIHR_IRSC, and DoD Prostate Cancer Research Program Idea Development Award!
Finally, we show in the paper, using real and simulated datasets that Freddie produces accurate and useful results without using the annotation database (e.g. ENSEMBL).
One of the most exciting parts of Freddie is its clustering/error-correction step which clusters long-reads while facing two challenges:
1) not knowing a priori how many isoform clusters exist
2) needing to account for correctable errors that result from sequencing
ALT In Freddie, we formulate the clustering step as an optimizing problem that we coin MErCI (Minimum Error Clustering into Isoforms). As a result of the segmentation step, long-reads are summarized as segments. The clustering step runs in rounds. In each round reads are clustered into a "recycling" bin or an isoform bin based on their similarities/differences: a read in the recycling bin incurs a constant cost while ones in the isoform bin incur cost that depends on how many errors they contain. Recycling bin reads are used as input to next round of clustering.
Freddie employs a multi-step approach to detect isoforms where each step is designed with high error-prone long RNA reads.
ALT Freddie has three main stages: partitioning, segmentation, and Clustering/error-correction. After partitioning, different subsets of the input data can be processed in-parallel for faster computation.
Our work on Freddie, an annotation-indepented isoform discovery using long-reads tool, is finally published at @NAR_Open! Check the open access code and paper:
📄: academic.oup.com/nar/advance…
💾: github.com/vpc-ccg/freddie/
🐍: conda install -c bioconda freddie
This work could not have been possible without our close collaboration with @RyanFlannigan00 lab (including Meghan Robinson!) and Cedric Chauve
And of course, all the intuitional support we got!
@VCHResearch @VanProstateCtr @CRSNG_NSERC @HlthResearchBC @UBC_CS @SFUMath
In our manuscript, we show that scTagger is highly accurate and is much faster than the alternative methods and demonstrate that on both real and simulated datasets.
