The moka-pot-coffee image is from https://bsky.app/profile/nrennie.bsky.social/post/3mjukrwxyxs2y
2026-08
Most mammalian retinae (视网膜) only contain 2 types of cones (视锥细胞，一种感光细胞) (dichromats), sensitive to short wavelengths (S-cones, maximally sensitive near the blue end of the visual spectrum) and medium wavelengths (M-cones, maximally sensitive to green wavelength light). Humans, as trichromats, have an additional L-cone, which is sensitive to long-wavelength light at the red end of the visual spectrum.
Physiology, Color Perception, 2022
… for us the human eye only has three different colour receptors, and so we can summarise the perception of any colour with just three numbers.
ggplot2: Elegant Graphics for Data Analysis (3e)
The absence of validated immune correlates of protection (against tuberculosis) and the uncertain predictive value of animal models means that candidate vaccines need evaluating in large, expensive efficacy trials.
DRT3b forms a hexamer and uses active-site-adjacent residues as deoxyadenosine and deoxycytidine gates to enforce alternating nucleotide addition, representing a unique example of amino-acid-templated DNA polymerization.
… We defined three distinct phenotypes based on our observed (intracellular) doubling times relative to literature values: normal growth (16-26 h), slow growth (>=24 h) and fast growth (<=16 h).
Unexpectedly, we discovered fast-growing intracellular Mtb populations with doubling times below 10 hours.
2025-07
Antibody function is governed not only by antigen specificity, conferred by the antigen-binding (Fab) domain to neutralize pathogens but also by the antibody constant (Fc) domain, which recruits complement and engages Fc receptors (FcRs) on the surface of immune cells.
Immunity. 2025. PMID: 40449485
Any given MHC molecule can bind numerous different peptides, and some peptide can bind to several different MHC molecules. Because of this broad specificity, the binding between a peptide and an MHC molecule is often referred to as “promiscuous”.
Kuby immunology 2018 p533
…studies use a cross-validation approach, in which one chunk of a data set is reserved for training and another for testing. This can lead to issues such as overfitting, meaning that an algorithm performs well on data similar to the information that the model has encountered before but does poorly on disparate data. “The problem with cross-validation is that it tends to provide fairly optimistic results,” Madabhushi explains. “The cleanest and best way to validate these models is through external, independent validation, where the external test set is separate and distinct from the training set and ideally from a separate institution.”
Trained immunity in alveolar macrophages can lead to damaging lung inflammation, confirming the importance of context in this phenomenon.
eLife. 2025. doi: 10.7554/eLife.107787
Whether it’s a sales skill, writing skill, musical talent, preaching ability, inventiveness, food production, cleaning, or childcare, these are all value-creation skills. Few people possess all these skills, and no one is the world’s best at all of them. That’s why most people specialize and outsource the rest.
Actually you have 22 pairs of chromosomes numbered 1 through 22, and if you’re a woman, you have a pair of X chromosomes, for a total of 23 pairs. If you’re a man, you have one X and one Y chromosome instead of two X’s, so the 23rd “pair” isn’t technically a pair at all, even though some parts of Y are identical to some parts of X.
It is important to be aware that most common image/photo editing software applications don’t care about preserving pixel values. Only the appearance really matters. Therefore if you adjust the brightness/contrast of an image in such software you can easily make the image unsuitable for later quantitative analysis by compromising the data.
In practice, magnification is then not terribly meaningful: the pixel size is the crucial value needed to make measurements in physical units, and to compare the resolution across images.
Writing compels us to think — not in the chaotic, non-linear way our minds typically wander, but in a structured, intentional manner.
In QuPath, as in life, it is neither easy nor wise to rewrite history.
I am a strong believer that depth of data can be just as powerful as scale of data. Deep analysis of small groups of humans, and even individuals, can lead to novel biological insights and principles.
Energy and initiative matter more than talent.
2025-08
Quanlity before quantity.
The testis is known for its immune-privileged status, maintained by the blood–testis barrier, which prevents autoimmune reactions against spermatogenic cells.
Med Sci (Basel). 2025. PMID: 40843751
… doesn’t yet have the thick skin to shrug off the hostility.
Good ideas come from having lots of experience. And experience comes from having lots of bad ideas.
2025-09
the rapid growth and accumulation of sequence data have outpaced efficient analysis capabilities.
Clear lines of logical argument, precise definitions, often can only be composed in writing. Paper and pencil (or the word processor) are extensions of our brains.
… image analysis may be faster, more reproducible, and more quantitative than visual assessment.
… story-telling is easier than science.
As far as the computer is concerned, each pixel is really just a number and the full image is a 2D matrix of these numbers: the pixel values … In science, the pixel values are crucial - how they are displayed is not.
2025-10
UC wins 5 Nobel Prizes in 3 days — and sets a new world record.
Great to know that we are not alone in this effort.
Joseph Arboleda-Velasquez on LinkedIn
The naked mole-rat/裸鼹鼠 (Heterocephalus glaber) is an extraordinarily long-lived rodent with a maximum life span of ∼37 years.
Science. 2025. DOI: 10.1126/science.adp5056
The naked mole rat (Heterocephalus glaber) is the longest-lived rodent, with a maximum life span of nearly 40 years.
Science. 2025. DOI: 10.1126/science.aeb6125
We complement each other very much.
Peter Howitt, laureate of 2025 Nobel Prize in Economics Sciences, on X
The more variability you capture in the training data, the more accurate the classifier will be.
MicroscopeRA on forum.image.sc
At more than 200 years, the maximum lifespan of the bowhead whale exceeds that of all other mammals.
2025-11
… in most cases you’ll have to train the projects on multiple images, representative of the whole dataset.
… skepticism belongs in science - particularly in image analysis - and it’s always worth testing things from multiple angles, just in case.
Introduction to Bioimage Analysis
The first thing you need to figure out is why the intensity changes across images. Is it something real about the biology, or inconsistent staining, or changing acquisition protocols? Also, is it that some images, as a whole, are brighter than others, or are there regional changes throughout a slide?
… commercial kits available for quenching AF (autofluorescence) … Good quality sections with clean staining and well-scanned images is by far the most immportant part of image analysis.
Consider that perhaps intensity based gating (even with correction) is not going to get you there… probability based object classifier, or Deep Learning based phenotyping … can take into account more than just intensity (eg. textural features, patterning, puncta, polarisation), and in fact be more robust against changes field-to-field and image-to-image… be comfortable with an error rate (however small). Given the complexity of multiplexed tissue image analysis, there will be errors.
… image analysis is not for perfectionist or you’ll never ended: just be rigorous with good controls.
2025-12
… if the function of two gene products completely overlaps then, throughout the course of evolution, one of the genes will likely accumulate mutations to the point of loss-of-function.
M. tuberculosis genome retains redundant functions at multiple levels that allow the bacilli to adapt to extremely heterogeneous environments in the human host.
PLoS Pathog. 2025. PMID: 41348814
When working with an image in ImageJ, I almost always have at least one duplicate lingering around for when things inevitably go wrong.
Introduction to Bioimage Analysis
R is widely used in statistical computing, data analysis, and bioinformatics. A key contributor to its success in bioinformatics and computational biology is the open-source project Bioconductor.
bioRxiv. 2025. doi: https://doi.org/10.64898/2025.12.22.695980
Almost all interesting image analysis involves filtering in some way at some stage.
2026-01
The Greenland shark is the longest-living vertebrate on Earth, with an estimated lifespan of up to 400 years.
Nat Commun. 2026. PMID: 41491153
I think that’s where QuPath has the most value: not trying to do everything for everyone, but trying to do some things well … and expand its horizons over time.
2026-03
arXiv, the granddaddy of preprints servers.
You can publish a paper with wrong data. You can’t make a working drug with a wrong experiment.
Citation
@online{wu2026,
author = {Wu, Kang},
title = {Some Good Sentences in Biology},
date = {2026-08-27},
url = {https://iamhealthy.github.io/nutrition/maybe_good_sentences_in_biology/},
langid = {en}
}