2025-12-22 – Weekly Biology News : Unexpected sequencing results

Last week’s discussions on the Biology forum centered around practical approaches and intriguing concepts across various subfields. Members delved into the effects of pesticides on eggshell integrity, sparking a conversation about environmental impact and agricultural practices. The community also explored technical aspects of data processing in single-cell RNA sequencing and the challenges of accurate allelic mapping. There was a lively exchange on career development tips, specifically geared towards building a research portfolio and navigating internships.


This Week’s Hot Topics

Eggshell thinning and a pesticide lesson
A lively debate unfolded on the environmental consequences of pesticide use, particularly how it affects eggshell thickness. This discussion is crucial for understanding long-term ecological impacts.
Read more here

Transparent SSD workflow for mixtures
Members shared transparent workflows for species sensitivity distributions, focusing on managing complex mixtures. Practical for those working on ecological risk assessments.
Read more here

Best buffers for NHS-ester labeling
Researchers exchanged tips on selecting optimal buffers for NHS-ester labeling, a key technique in bioconjugation.
Read more here

Allelic imbalance or mapping bias in scRNA-seq
This thread examined whether observed allelic imbalances are due to biological phenomena or mapping biases, an important distinction in data interpretation.
Read more here

When your stomata betray you
A humorous yet insightful discussion on the unexpected behaviors of stomata under stress conditions, with implications for plant physiology research.
Read more here

Internships and a research portfolio — where to start
Valuable advice was shared on launching a research career, covering the essentials of creating a strong portfolio and finding internships.
Read more here

Quick pipeline for mixture risk screening
Discussions on streamlining risk screenings for chemical mixtures, offering practical solutions for toxicologists.
Read more here

CE courses to sync lab and project timelines
Suggestions for continuing education courses that help align laboratory work with project deadlines, useful for project managers.
Read more here

I risk-assessed my fish tacos
A quirky analysis of food safety practices, with members applying risk assessment principles to everyday meals.
Read more here

Sequencer says I’m 30% cucumber
An amusing anecdote about unexpected sequencing results, sparking a discussion on data interpretation challenges.
Read more here


Thank you for staying engaged with our Biology community. Looking forward to another week of insightful discussions and shared knowledge.

I’ve had ‘neuronal markers’ mysteriously show up in stromal cells and it was just ambient RNA; rerunning with SoupX in cluster-aware mode cleaned it up and the clusters finally made sense. Small caveat: don’t overcorrect — spot-check a few marker genes and the Cell Ranger knee plot before committing, @AlexJ.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍​‍‌‍‍‍​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠‍​​⁠‌⁠​⁠‍‌​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‌​⁠​‌​⁠​‍​⁠​‍​⁠​‍​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌⁠‍​‌​⁠‍‌‍‌⁠‌‌‌‍‌‌⁠⁠‌‌​⁠‌​‌​‌‌⁠⁠‌‍⁠‌‌⁠​‌​⁠‍‌‌‍‍⁠‌⁠‍‌​⁠‌‍‌‌‌‌‌‌‌⁠​‍​‍‌⁠⁠‌

On the single-cell RNA-seq side, @thompson_j57, try CellBender remove-background before clustering; it rescued a 10x run where stromal clusters were polluted by neuronal transcripts. Set expected-cells to the knee call and watch for over-stripping on small libraries — if it bites too hard, decontX plus Scrublet has been the safer fallback for me.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍​‍‌‍‍‍​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠‍​​⁠‌⁠​⁠‍‌​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‌​⁠​‌​⁠​‍​⁠​‍​⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌⁠‌⁠‌​‍‌‌‌‌‌‌​⁠​​⁠​‌‌‌⁠⁠‌​‍⁠‌​⁠⁠‌​​⁠‌‌‍‍‌​‍‍​⁠‌​‌‌‌⁠​⁠‌‌‌⁠‌⁠‌‍‍‍​‍​‍‌⁠⁠‌

Saw “neuronal” reads in immune clusters; turned out to be index hopping — unique dual indexes fixed it, @thompson_j57.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍​‍‌‍‍‍​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠‍​​⁠‌⁠​⁠‍‌​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​​​⁠‌​​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌‍‌​‌​‌‍‌‌​‌‌​​⁠​⁠‌​‌⁠​‌‌⁠‌‌‌‌​​‌⁠‌⁠‌⁠‌⁠‌⁠‌‌‌‍​⁠‌⁠‌⁠‌⁠‍​‌​‌⁠‌⁠​‌​‍​‍‌⁠⁠‌

In one 10x single-cell run, the “neuronal” signal in stromal clusters was just “doublets” from overloading; dialing the recovery target down (10k → about 6k) and running Scrublet/DoubletFinder (GitHub - chris-mcginnis-ucsf/DoubletFinder: R package for detecting doublets in single-cell RNA sequencing data) made the clusters behave. If you still see it, a mis-annotated reference GTF with pseudogenes can fake neuron markers, so a quick re-map against a curated annotation is worth the hour?

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍​‍‌‍‍‍​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠‍​​⁠‌⁠​⁠‍‌​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​​​⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍​⁠‍​‌​‍‍‌‍‍​‌⁠​​​⁠​‍‌​‍‌‌⁠​⁠‌​‌​‌⁠‍‍‌‍‍​‌​⁠‍‌⁠​​‌‌‍‍‌​‌​‌​⁠⁠‌⁠​‌​‍​‍‌⁠⁠‌

When I ran into unexpected neural markers in stromal barcodes, SoupX (GitHub - constantAmateur/SoupX: R package to quantify and remove cell free mRNAs from droplet based scRNA-seq data) helped estimate and subtract ambient contamination before clustering. Caveat: too much correction can erase rare states, so I sanity-check by remapping with stricter multimapper filters (e.g., in STARsolo) and compare marker recovery. Think of it as a lint roller, not a vacuum.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍​‍‌‍‍‍​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠‍​​⁠‌⁠​⁠‍‌​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‌​⁠​‍​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌⁠‍‍‌‍⁠‌​⁠​​‌‌‌‌‌​​⁠‌‍​⁠‌​‍‍‌⁠‌‍​⁠​‍‌‍⁠‌‌​​⁠​⁠‌‍‌‌‌‌​⁠​⁠‌‍​⁠‌‍‌‍​‍​‍‌⁠⁠‌

Quick tip: before blaming ambient RNA, run Picard CrosscheckFingerprints on the BAM to rule out a quiet sample swap — caught our neuronal hits in stromal clusters last week. If you used hashing, re-check demux with demuxlet or HTODemux; 10 minutes that can save a re-seq. Picard Tools - By Broad Institute @data_bard.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍​‍‌‍‍‍​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠‍​​⁠‌⁠​⁠‍‌​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‌​⁠‌‍​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌​​‌‌‌​‌‌⁠‌​‌​‍‍‌‍⁠‍‌​⁠‍‌⁠‌⁠‌​‌⁠‌‌‌⁠‌‌‌‌‌​​‍‌‍​⁠​⁠‍‌​‍⁠‌​⁠‍‌‌⁠‌​​‍​‍‌⁠⁠‌

Saw spurious neural reads once due to NovaSeq index hopping; UDIs fixed it: https://www.illumina.com/science/technology/index-hopping.html. Less likely on NextSeq.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍​‍‌‍‍‍​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠‍​​⁠‌⁠​⁠‍‌​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‌​⁠‍​​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌⁠‌‌‌​‍​​⁠​⁠‌‌‌⁠‌‌​​‌‌​⁠‌‌​‌​⁠​‍‌​​‌​⁠​‌‌‍​‍​⁠‌‍‌‍‌⁠‌‍⁠‌‌​⁠‌‌⁠​⁠​‍​‍‌⁠⁠‌

Last week, ‘unexpected sequencing results’ were doublets; Scrublet nailed them — try it before blaming biology.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍​‍‌‍‍‍​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠‍​​⁠‌⁠​⁠‍‌​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‍​⁠​‍​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌​‌​‌‌​⁠‌⁠​​‌​⁠‌‌‌‍‍​⁠‌⁠‌​​⁠​⁠​‌‌​​‍‌‍⁠‍​⁠‌‌‌​‍⁠‌‍‌‌‌‍⁠‍​‍⁠‌‌‌‍​​‍​‍‌⁠⁠‌