n this study, we introduced a method for representing transcriptome components usinglow-dimensional Raman LDA axes based on Raman-transcriptome linear correspondence (Fig.4).
it would also be valuable to see which Raman wavenumbers drive the LDA axes to see if they make conceptual sense and are not just noisy artifacts. by eye, the representative spectra in fig 1c are very similar, so it would be good to see that the differences between the strains are driven by real differences in peaks
The transcriptomes of the nine S. aureus strains were ob-tained from the public database
were the transcriptomes (or at least a subset) validated in-house? some more details about how the transcriptome data was used would be very beneficial to the reader.
We compared the performance ofRamanMAE-based smoothing with the Savitzky-Golay algo-rithm (which is sensitive to polynomial order and windowsize choices) and found that RamanMAE provides equivalentor slightly better similarity between reconstructions from thelow and high SNR spectra
from the plots, it seems like using S-G for denoising is comparable to RamanMAE. it would be good to know what the advantage is of RamanMAE as a much more computationally intensive algorithm
Figure 1.
i am confused how the 2D representation (image) is actually created. does the wrapping go patch-by-patch or by the entire width of the image? perhaps an illustration could be useful.
evertheless, even for tetrapep-tides, the Recall@1 for retrieval and generation can still reach 68.03% and 48.18%,respectively
Does the model actually predict the correct sequence of AAs as opposed to just the AA content? If so, that is amazing, but I would be very surprised if a short peptide e.g. "RACR" had a different spectrum than a peptide containing the same AAs but in a different order e.g. "RRAC". If this is actually the case, it may be very informative to actually show the predicted spectra!
Table 1
how does this performance compare to other techniques of predicting sequence from spectra? e.g. a simple linear model that tries to optimally fit known AA spectra to the unknown spectrum?
(A) and (B)
Great informative figure but the colorscheme is slightly confusing. perhaps dataset titles do not need to be colored to avoid confusion with the colors of the models in the plots?
(E) re-ranking module for refin-ing retrieval and generation results. It initially filters candidates by chemical formula (if available),then uses a pre-trained molecular encoder to score them against the query spectrum. High-scoringcandidates are finally selected as output
the re-ranking module seems to look at existing chemical-spectral pairs and rank ones within the training set as more likely (higher score) and those outside the training set as less likely (lower score). if that is the case, does it negatively impact performance when inferring on out-of-distribution data? and what would be the performance of this module alone (i.e. without the encoder/decoder models altogether?
Fig. 2
great figure! very clear and descriptive.
Graph of accuracy of CNN when the test set is perturbed at various28wavenumbers. Sharp decreases can be seen with perturbation at bands associated with (i) adenine, (ii) thymine, (iii)29aryl, (iv) phosphate, and (v) carboxyl.
this is a nice way to identify which peaks contribute to the separability of the species, especially when it is not clear by eye. As an alternative, it may be possible to simply do an ANOVA across the species at every datapoint.
Error bars indicate standard deviation of 1599 cm-1 peak, which is representative of8standard deviation for all peaks.
i understand what this is showing but it seems like an unconventional way to show this kind of data. The shading colors make it seem like there is some kind of continuous concentration sweep from 0 to 150.
We preferentially enhance Raman signal from bacteria in19wastewater using positively-charged plasmonic gold nanorods (AuNRs) that electrostatically bind to the20bacterial surface
will this result in false positives due to other benign bacteria that may be present in the wastewarer that is still negatively charged or other non-bacterial cells?
We note that the data had to be corrected for slow magnetic field (B0) drifts caused by thetemperature-dependent magnetization of the permanent magnet.
how much does temperature rise from room temp? does it reach steady state?
Optical widefield magnetic resonance microscopy (OMRM
This is a very impressive technique. Are there any limitations to the types of substances that can be assayed/detected? The demonstration here is with water, but it would be very interesting to see the technique's detection limits when applied to other compounds.
The future of benchmarking for deep learning methods, as we move toward more complex multidimensional taskssuch as co-folding, requires different measures to assess leakage and difficulty, whether for protein-protein interactions(PPIs), protein-ligand interactions (PLIs), or protein-nucleotide complexes
is one possible solution (or at least mitigation strategy) to include some kind of confidence score in PLI prediction scores?
This is also particularly crucial for taskswith even more limited data, such as those involving covalent bonds and modified residues
does it follow from this that training PLMs on smaller, more uniformly sampled and diverse datasets could lead to better generalization? or is this overfitting an unavoidable feature of larrge models?
We demonstrate that the performanceof current approaches strongly correlates with the similarity to their training data, regardless of the metric used todefine success or the subsets considered
is there a metric that can be used for subsequent models that can give users insights into how problematic overfitting is in that particular model? something that takes the training set similarity into account.
Figure 3:
very clear and intuitive figure.
In the first round of design, a maximum edit distance cap of 6 edits444from the lead is enforced. In the second round, this cap is increased to 8, and in the445third round, it is increased to 12.
This seems still quite close in sequence space to the original sequences -- how do the results from this method compare to other protein engineering efforts on the same targets? are there cases where the same residues were mutated?
DCS uses a likelihood under a joint density of statistical properties, includ-431ing log-probability under a protein language model, and sequence-based properties432like hydrophobicity and molecular weight, calculated with BioPython
it seems from a first pass to be a little bit circular to use OOD detection on PLM-generated sequences. Presumably PLMs have "learned" various aspects of what makes a good (in-distribution) protein and already incorporate (implicitly) some concept of hydrophobicity, MW, etc. In other words, it would be interesting to see for cases of major disagreement between the generative model and the OOD-detection model, which one is correct?
3
how was this metric chosen?
Our results demonstrate the powerful generalization capabilities of LitL to perform102antibody design across diverse antigen targets and epitopes, without human inter-103vention, while producing real therapeutic antibodies that are viable candidates to104progress in the drug discovery pipeline.
how does this connect to ultimate success/failure and profitability in the clinic? i.e. if we apply this (powerful!) approach across all drug targets from now on, will we get a significant boost in clinical performance? or are failures in the clinic caused by other factors not addressed by this method e.g. incomplete understanding of the disease states, selecting the wrong targets, choosing the wrong patient population etc?
Supplementary Fig. S
what do the different colors signify in the plots?
incorporating only mutations found in previously stable sequence
does that mean that the GA will not consider mutations not encountered in the binding affinity datasets?
Designs express and bind at consistently high rates (> 85%), comparable to that of singlepoint mutants.
it would be interesting to see a naive control, i.e. what is the average expression and binding rate if you just make N point mutations at random?
66 pM, exhibiting a near 50-fold improvement
Very impressive!
DyAb performance on the regression task for design sets are shown in Supplementary Fig.S3
from S3a, it looks like DyAb is not very predictive with the lead A dataset, but performs much better on the others even though they have equal/fewer data points. Any idea on why this is?
Only cultures where neurons were evenly distributed across the entire coverslip
part of a sentence missing
for single
typo?
As these experiments did not include synaptic blockers, the true indicatorkinetics are likely faster than measured
true; adding synaptic blockers improved kinetics but also dramatically increased day-to-day repeatability of the GCaMP-series sensors. this would probably be worth doing for a large-scale screen.
CTZ
Acronym not previously defined
(A)
In A and B, is the stimulus the same or increasing in the number of pulses? Trying to understand why the response seems to grow over time.
(A)
For maximum clarity, it may be useful to use d prime or SNR as the key metric for comparison because both metrics are designed to measure what most users would care about: how detectable is the Ca2+ signal regardless of baseline brightness.
diffusion of furimazine (Fz) m
it may be helpful to contextualize the role of furimazine before this, as it may not be familiar to readers without BL assay experience
(C)
Stylistic note: the label makes it seem like "CaBLAM" and "GCaMP8s" in D is a trace on the plot along with ROI1,2,3. Maybe it can be left out?
A
what is the box on the bottom right in A? Presumably it is the camera? A label would be helpful.
Large-field tiled acquisitions highlight nuclei and clear cellboundaries, enabling the simultaneous observation of numerous cells.
how does this compare to just using standard brightfield imaging or other label-free techniques (e.g. DIC)? are there any features that quantitative phase picks up that other techniques miss?
The precipitous drop in RI valuesmay be attributed to changes in cell density and volume as cells spread out,
Is it possible to determine computationally whether the drop in RI is only attributed to density change or something else?
acoustic focusing from an attached pi-ezoelectric transducer to confine cell motion to the center ofthe capillary
a citation describing acoustic focusing may be useful. is it known how well the particles are confined using this technique?
Conclusion
very impressive technology!
-865, nac Image Tech-nology
i may have missed it but what is the CMOS camera used for in the study?
depending laser out-put
typo: depending on laser output
0.26 – 3.2 mW/μm
this seems like a large amount of power, is it possible to theoretically determine the temperature increase that this light could induce in a transiting object?
RepresentativeEGFP and mStayGold nanocage photobleaching images. Images are scaled to the same absolute intensities
these are not the same nanocages being imaged at the three different timepoints, correct? would it be possible to image the same immobilized nanocages over time?
we present a new method to compare fluorescent proteins on a molecule-by-molecule basis inphysiological conditions independent of relative expression levels or ratio measurements to other fluorescentproteins
As a tool to evaluate FPs in the future, what does this approach provide above what can be learned with absorbance/spectroscopy measurements performed on purified FPs?
omparable peripheral cellregions of D-Mannitol-treated RPE cells expressing the indicated FP nanocage fusions
In the last panel (E138D), it looks like there are two dim nanocages and the rest are brighter. What could be the source of the variation?
Thus, each nanocage carries sixty FPs (Fig. 1b) allowing comparison of the absolute fluorescence intensitybetween individual nanocages with different FPs
this is a very clever way to benchmark FPs that ensures a direct head-to-head performance comparison. By making comparisons between individual nanocages, this technique largely avoids confounds introduced by variations in expression level and cell-to-cell variability.
hen γ, whichsets the cell membrane tension, is low, i.e., the cell is soft,deformations are not very costly and can occur more eas-ily.
Is it possible to measure membrane tension using Brillouin microscopy or AFM? Will there be a direct linear correlation between the measured tension (or maybe Young's modulus of the cell) and the tendency for bistability?
Evidence suggests that cell stiffness is apotential biomarker for cell malignancy with cancer cellstending to be softer [45–48]
Are there other biophysical properties of the cell that this technique is capturing that are not captured by cell stiffness alone? It would be really interesting (and clinically useful) if the bistability property was informative of the cell's propensity to become malignant
FIG. 3.
Is there an optimal enclosure geometry that maximizes the differences in behavior?
Within our model, we can switch cells between showinglimit-cycle dynamics and bistability either by changingphysical parameters like cell size or tension, or by chang-ing polarity dynamics like the size of protrusion fluctu-ations.
Would it be beneficial to do an e.g. RNAi screen to determine how different genes contribute to motility?
Further experimentalstudies will also be necessary to validate our suggestionof cell properties as predictors of cell motility.
Do you think there could be clinical applicability? e.g. if you optimized the geometry to try and maximize the difference in motility b/w likely cancerous and non-cancerous cells, could this technique be used to predict pathology where other techniques may fail?
While nuclear dynamics could be mod-eled with another phase field [9], we neglect it here forsimplicity and claim that cell center of mass is a fair ap-proximation of cell nucleus for most of the morphologieswe observe.
If a cell nucleus (or another subcellular structure) is close to the edge of the cell, would that impact cell polarity calculation?
A beter strategy would be to interleave epochs of intense (> 10 mW/mm2) illumina�on withepochs of darkness. Similarly, for voltage imaging of large samples (e.g. an en�re mouse heart), theexcita�on intensi�es may be low, leading to a loss of voltage sensi�vity.
That's a very valuable insight! Do you think there would be any benefit in simultaneous 1P and 2P illumination to increase SNR?
For interleaving stimulation, is the idea to use the kinetics from Fig 2d,e to determine the min period that can be used to illuminate the same spot?
he complex photophysics of the FRET-opsin GEVIs suggest that future protein engineering efforts shouldbe accompanied, at a minimum, by a quan�fica�on of intensity-dependent voltage sensi�vity. Aninteres�ng avenue for future explora�ons would be to determine the photocycle basis for the intensity-dependent changes in voltage sensi�vity and voltage step-response waveforms shown in Figs. 1 and 2..CC-BY-NC-ND 4.0 International licenseavailable under awas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is madeThe copyright holder for this preprint (whichthis version posted April 2, 2024.;https://doi.org/10.1101/2024.04.01.587540doi:bioRxiv preprint
Is it conceptually possible to engineer out the S2 excited state or change its excitation wavelength?
This observa�on shows thatstatements of FRET-opsin voltage sensi�vity are only meaningful if illumina�on intensity is specified
That's very good to know!
Figure 2
Maybe I missed it but the "I" and "II" labels in a,b,d,e are not explained in the legend. I assume I = low-power light regime and II = high-power?
b) Fluorescence traces (le� axis, detrended ΔF/F; right axis, F) of from 1P (top) and 2P(botom) epochs of a single recording
In the 2P panel it looks like the spike amplitude paradoxically decreases as you increase laser power -- is that due to photobleaching?
suppressed LC
wording: "stGtACR2-expressing neurons are suppressed in the LC"?
To ensure a smooth transition from therestraint stress to hot plate testing, we outfitted mice with a head-fixation bracket around bilateralfiber optic implants above the LC
it would be great to see a picture of the bracket!
Figure
Minor legibility note on panels E and F: it is difficult to distinguish the injured vs non-injured mCherry and stGtACR2 line markers, especially since there is also blue shading on the "stim"
Altogether, these resultssuggest rescued mu opioid receptor function may be a therapeutic target for the treatment chronicpain
could be promising but challenging from regulatory point of view given the opioid crisis.
(b)
Is it accurate to say that the retardance and orientation of infected cells was not different from uninfected? it may still be useful to plot for completeness. it may also be useful to note what kinds of investigations this unified fluorescence/label-free system allows that would not be possible (or be very challenging) with two separate systems. Overall, beautiful, useful paper!
Our iterative alignmen
It would be interesting to learn from a user's perspective how long the initial assembly takes and how long day-to-day routine maintenance takes
Prediction of protein-protein binding sites using embeddings
it would be interesting to see more details here about which PPIs are predicted well/not well by SSEmb vs other models. Is a particular type of PPI consistently missed? It would also be interesting to identify what causes false positives.
otentially novel regulators of mTORC1
it's very cool that previously unknown genes contributing to a pathway can be identified!
a-b.Comparison of feature embedding methodologies based on median AUC of binary classification of KOfrom WT for each genetic perturbation.
It looks like the majority of the genes have AUC ~= 0.5; what is the interpretation of that? Does that mean that most gene KOs tested do not exhibit a phenotype distinguishable from wild-type?
he field of view images are then cropped around thecentroids of each of the segmented nuclei and masked by the corresponding cell segmentation mask tocreate tiles with a single cell in context
are pixels that are outside the mask painted with zeros on all channels?
Fig. 2
It would be interesting to know what is the limit of detection? i.e. the smallest voltage fluctuation that could be reliably detected?
cell types
Another interesting cell type could be insulin-secreting beta cells. A similar ratiometric sensor was developed and tested in those cells in 2017: Schifferer, Martina, Dmytro A. Yushchenko, Frank Stein, Andrey Bolbat, and Carsten Schultz. “A Ratiometric Sensor for Imaging Insulin Secretion in Single β Cells.” Cell Chemical Biology 24, no. 4 (April 20, 2017): 525-531.e4. https://doi.org/10.1016/j.chembiol.2017.03.001.
Fig. 5. Fluorescence–fluctuation correlation analysis uncovers strong electrical coupling among HEK293T and A375cells and weak coupling among MCF7 cells
This is a very cool result and it would be very interesting to see actual videos of the electrical coupling in action. e.g. would it be possible to see millisecond-scale membrane activity actually spreading via the gap junctions? Similar to what was done in Fig 3c of https://doi.org/10.1038/nmeth.3000?
Fig. 5. Fluorescence–fluctuation correlation analysis uncovers strong electrical coupling among HEK293T and A375cells and weak coupling among MCF7 cells
This is a very cool result and it would be very interesting to see actual videos of the electrical coupling in action. e.g. would it be possible to see millisecond-scale membrane activity actually spreading via the gap junctions? Similar to what was done in Fig 3c of https://doi.org/10.1038/nmeth.3000?
cell types
Another interesting cell type could be insulin-secreting beta cells. A similar ratiometric sensor was developed and tested in those cells in 2017: Schifferer, Martina, Dmytro A. Yushchenko, Frank Stein, Andrey Bolbat, and Carsten Schultz. “A Ratiometric Sensor for Imaging Insulin Secretion in Single β Cells.” Cell Chemical Biology 24, no. 4 (April 20, 2017): 525-531.e4. https://doi.org/10.1016/j.chembiol.2017.03.001.
Fig. 2
It would be interesting to know what is the limit of detection? i.e. the smallest voltage fluctuation that could be reliably detected?
It is the centre of mass that is used as the x/y targetcoordinates for the pipettes, so as to patch the centre of thecell, while the z-coordinate comes from the plane at whichthe cell is most in focus.
Do you have situations when the cell moves out of the way and you have to take the pipette out and try again?
After each trial, pipettes are automatically cleaned in anenzymatic solution and then returned to their initial positions,allowing for reuse,
was wondering how often you guys encounter clogging from debris in the pipette after cleaning? we saw this happen a lot especially if the pipette was close to vertical position
After each trial, pipettes are automatically cleaned in anenzymatic solution and then returned to their initial positions,allowing for reuse,
was wondering how often you guys encounter clogging from debris in the pipette after cleaning? we saw this happen a lot especially if the pipette was close to vertical position
It is the centre of mass that is used as the x/y targetcoordinates for the pipettes, so as to patch the centre of thecell, while the z-coordinate comes from the plane at whichthe cell is most in focus.
Do you have situations when the cell moves out of the way and you have to take the pipette out and try again?
Fig. 1
Can the model run at inference time with only a single image or z stack as input? Also will the model be publicly available at some point? Great work!!
Fig. 1
Can the model run at inference time with only a single image or z stack as input? Also will the model be publicly available at some point? Great work!!