Pi-Chuan Chang

Technical lead / Eng Manager of DeepVariant and DeepConsensus.

Currently at Google; previously at LinkedIn, AltSchool, Google, Stanford

[Yossi Matias @ymatias]

A milestone in Somatic variant detection: New research introduces DeepSomatic, an AI tool that accurately identifies mutations in a cancer cell’s genetic sequence. This work, published in Nature Biotechnology in collaboration with @ucscgenomics, is a step toward making precision oncology more effective. The tool, and a new dataset, are open-source for the research community.
More in our blog by Kishwar Shafin and Andrew Carroll: https://t.co/DFcfvGFKEW

Read the full Nature article: https://t.co/kQJuANOL3D

[NVIDIA Healthcare @NVIDIAHealth]

How fast is fast?

⚡ Over 14x faster runtimes with pangenome-aware DeepVariant 1.9 + Giraffe on four NVIDIA RTX PRO 6000 GPUs vs CPUs. Plus, DeepSomatic long read and WES support, quantMode GeneCounts for STAR, and Mutectcaller mitochondrial mode.

What could this mean for your genomic workflows?

[Andrew Carroll @acarroll_ATG]

Release of DeepVariant 1.8. Large speed improvement (~67% faster) via small model for easy sites. New Pangenome-aware option. Reduces error by ~30% for vg-mapped WGS ~10% for BWA WGS ~5% BWA exome. New config for custom model users, see release notes 1/3
https://t.co/TQmQElqAOR

[Pi-Chuan Chang @pichuan]

Initiated in 2023 by @marianattestad and @_beenkim, this project has been a collaborative effort.

For over a year, Atilla and Yuchen dedicated 20% of their time to working with @_beenkim , @acarroll_ATG and myself on this project.

It has been a fun exploration!

[Google AI @GoogleAI]

When we train deep learning models for genomics, what do they learn? To help answer this question, we examined the DeepVariant model to determine what insights it has developed, and we discovered some surprising concepts embedded within. Read more at https://t.co/SSUh4EdJVo

[Kishwar @kishwarshafin]

Transformer-based polishing approach DeepPolisher's manuscript is now live. @miramastoras polished 180 assemblies using DeepPolisher for the next human pangenome release. Collaboration with @BenedictPaten @MobinAsri.

PM @acarroll_ATG and eng mng @pichuan.
https://t.co/lOmGI29ai1

[Pi-Chuan Chang @pichuan]

I have just finished mentoring for GetMAGIC this summer! https://t.co/dSRuwPqyow

We're looking for female mentors for middle and high school girls! Applications are now open for 2025 Spring (Jan-May).

Please see https://t.co/S8dMs7Ociu for more info.

[Andrew Carroll @acarroll_ATG]

Can you genotype a variant with only the information in the image below? We were surprised DeepVariant can get an F1 of 0.82 for Insertions with only MAPQ, vs F1 0.48 for random guessing.

Why? we investigate how DeepVariant uses signals in a new blog
(https://t.co/6w5Lp30V9V)

[Kishwar @kishwarshafin]

DeepSomatic (https://t.co/BqZkDNMv1E) preprint is out showing improvements in somatic variant calling in various platforms.

Work lead by @jiimiinpaark and @daniel_e_cook. Tumor-only lead by @pichuan. In collaboration with @acarroll_ATG, @MishaKolmogorov and @BenedictPaten. 🧵1/7

[Pi-Chuan Chang @pichuan]

#GivingWeek
Please consider donating to MAGIC (@STEMmagic )
==> https://t.co/O6YWfNqcpr <==

a non-profit that I’ve volunteered for since 2008. https://t.co/1kr6Qzb9ev. As I often said, “MAGIC volunteer” is the job I’ve held for the longest time. 🙂

[Pi-Chuan Chang @pichuan]

My 2023 summer mentee (through GetMAGIC) was Isha.

It was really fun working with her using microbit (https://t.co/WhvuRQOuLb) and writing Python to create a project!

It's great to see this being a spotlight on MAGIC's Instagram page! https://t.co/2bdMbdizbb

[Andrew Carroll @acarroll_ATG]

Release of DeepVariant v1.6.

Support for haploid regions, chrX/Y.
Workflow for Pangenome FASTQ-to-VCF.
Major DeepTrio improvements for de novo variants.
Models for CompleteGenomics T7, G400
Add NovaSeqX to training data

Release by @kishwarshafin

https://t.co/ShaxqDNTfX

[Andrew Carroll @acarroll_ATG]

Initial release of DeepSomatic, which identifies subclonal variants when given tumor and normal BAM files. Pre-trained models and case studies available for Illumina and PacBio. Development led by @kishwarshafin which built off a framework by @pichuan

https://t.co/YpB46RlhiB

[Pi-Chuan Chang @pichuan]

MAGIC (@STEMmagic) is looking for female mentors for for middle and high school girls. We're now open for applications for 2024 Spring (Jan-May).
Please see https://t.co/S8dMs7Ociu for more info.
I have been mentoring for MAGIC since 2008. Happy to chat more.

[Andrew Carroll @acarroll_ATG]

Happy to share this paper which provides more detail about the process we use to assign haplotypes to long reads on-the-fly, which enabled us to speed up the DeepVariant release at v1.4. Implementation by Alexey Kolesnikov, Lead for collaboration, writing, figures @kishwarshafin

[Get Magic @STEMmagic]

https://t.co/PPekxoRKhI

We are proud and thrilled to share this amazing blog from one of our MAGIC mentees, Sofia Sassone. Sofia's MAGIC mentor is Namrata GR Raut.

Sofia provides an in-depth look into nanoparticles and it's use on preservation of historical artifacts.

Super!

[Andrew Carroll @acarroll_ATG]

Excited to share this analysis of @ElemBio’s data (and their new Cloudbreak chemistry). We see accuracy advantages in Element data, especially at 20x-30x coverages and in homopolymers and repeats. Longer sequening inserts can further improve recall through hard genome regions.

[Pi-Chuan Chang @pichuan]

(btw, posting on Twitter reminds me of writing my old Perf packet.... cutting words takes more time than writing them :p)

[Pi-Chuan Chang @pichuan]

#education
I had mentoring sessions with two students (~8-9 grade).

It's fun to pair program. Interesting to see how people make mistakes and learn. Can difficult to figure out when to nudge them to think, when to just provide that information. Which is a learning opportunity!

[Pi-Chuan Chang @pichuan]

If you're interested in volunteering too, see https://t.co/S8dMs7Ociu :)

[Vivek Natarajan @vivnat]

Medicine is inherently multimodal.

Thrilled to share Med-PaLM M, the first demonstration of a generalist multimodal biomedical AI system with a stellar team @GoogleAI @GoogleDeepMind @GoogleHealth

Paper: https://t.co/oEMXZSW2bK