The 2025 Cold Spring Harbor Labs, Single Cell Analysis Meetings are here. This is the longest running single cell meeting, started in 2009. Please register and send an abstract!

with
streetslab.berkeley.edu
‪@saukaspengler.bsky.social‬

meetings.cshl.edu/meetings.asp...

I've started listening to vinyl again and enjoy going to used record stores.

~90% of them, I don't know the artists at all--but they were big enough to put out records. Then ~9%, I remember as huge but I doubt many people know them (Peter Frampton, anybody?).

This is prob the fate of our papers.

We used stacked transformers to learn a latent space that satisfy the 4-point cond. We defined a loss function that measures deviation from the 4-point cond. The prob is there are actually three sets of 4-point conds. Which set, depends on knowing the quartet tree topology.

8/n

There is a key theorem from 1970's by Peter Buneman called "4-point condition" that states that if a quartet fits a tree, their pairwise distances have to satisfy certain constraint equations.

*Note this important condition was derived by Peter for an anthropology problem!

7/n

All metazoan cells descend from a zygote forming a replication tree graph (cell tree). Cell differentiation lead to phenotypic diffs along the cell tree. But, cells that share phenotypic traits generally are not monophyletic (single origin), bc of var imp reasons too long to discuss here.

2/n

During these chaotic times, off to Seoul for RECOMB '25 to present our CellTreeQM (cell tree quartet metric) poster (Da Kuang and Guanwen Qui) to infer cell lineage trees from scRNA-seq using metric learning.

An overview (warning long) 1/n

A version of the paper is here:
arxiv.org/abs/2503.13925

Attending NHGRI Annual Training meeting in our home town of Philadelphia. This wonderful event with incredible platform talks from undergrads, PhD students, and postdocs, gives hope for a bright future if we can all work to keep the path forward.

Every little bit of resistance puts brakes on the abuse.