I am now sharing an extremely detaild new method invented by the OpenAI GPT-5 Pro model for one of biology’s oldest and most foundational techniques: Western blotting. I had the honor of learning it in 1989 in Switzerland from two of its original inventors, Drs. Harry Towbin and Theo Staehelin.

This AI-generated scientific method aligns with what @kevinweil announced today about starting OpenAI for Science, an AI-powered platform aimed at accelerating scientific discovery. It also supports what I have long argued: models like GPT-5 Thinking and Pro can accelerate the scientific process by orders of magnitude; this is another example!

It is important for non-scientists to realize that technology drives science. Without novel scientific techniques and methods, we cannot sustain accelerated discovery. This is especially true in biology, which is extraordinarily complex, with trillions of parts and dynamically interacting processes. Without AI we would have little hope of understanding all this complexity and solving biology, curing all diseases, or reversing aging in the coming decades, or even centuries.

This is why I've always been so exicted about AI and went all in several years ago after ChatGPT release, when it became clear to me the path to AGI has been opened. It is now time to recognize that this is not a hypothetical future any more; it is already here and getting better by the month, week, and day!

I should also point out that further development and proof-of-concept experimentations could turn this particular method into a patentable invention. However, I strongly believe that in the age of AI, such inventions should not be patented and should belong to all of humanity. In fact, our patent and intellectual property laws need to be substantially modified, or even replaced, to align with the reality that soon most inventions may be generated by AI, with human contribution increasingly focused on application and value creation. I digress; that is a deeper topic for another time :)

Before I share the method developed by GPT-5 Pro in the following thread, I would like to provide some historical background and explain the importance of the Western blotting method, which every biologist knows or should know. Here I am sharing a summary of GPT-5’s explanation; of course, anyone can check with Grok or other models for more detail:

What is Western Blotting?
Think of a cell extract as a soup full of different proteins. Western blotting does four simple things:
Sorts by size. An electric field pulls proteins through a jelly-like slab. Small ones run farther than big ones.
Copies the pattern to paper. The separated proteins are moved onto a thin membrane that holds them in place.
Finds your protein with a matching “key.” An antibody that sticks only to your protein is added, then a second antibody with a built-in label makes the spot light up or change color. The position tells you the protein’s size, and the brightness tells you roughly how much is there.

Who invented it, and when?
In 1979, two groups independently created the core method of protein “blotting” to a membrane after gel electrophoresis: Harry Towbin, Theophil Staehelin, and Julian Gordon in Basel, and Jaime Renart, Jakob Reiser, and George Stark at Stanford. Two years later, in 1981, W. Neal Burnette popularized the name “western blot.”

Why is it a pillar of biological methods?
Historically and practically it is one of the core lab methods for protein analysis. The reason is its specificity with size information. It tells you not just that a protein is present, but also its approximate size, which helps confirm identity, isoforms, and post-translational modifications. Indeed, Western Blotting remains one of the most commonly used protein assays across research fields and has been central to countless studies and workflows. Labs use it to verify antibody targets, confirm expression or knockdowns, and check pathway activation and use it in diagnostics (for example early HIV testing was dependent on this).

I will share the detailed method in the next thread. It is important to point out that this is a very sophisticated method, that took several prompts to validate. However, it can continue to be improved and optimized further and of course significant lab effort will be needed to develop it and troubleshoot it. This is a very detailed blueprint, which in itself is extraordinarily remarkable! In the third thread below I provide some critiques (also from GPT-5 Pro) and how to further validate and further improve it.

Because this is very technical it may be quite difficult for those not familiar with these type of methods to follow, so here is a scientific and layperson’s summary of the gist:

DSI-Seq (novel western blotting replacement)

Scientific summary of DSI-Seq: Digital Size-Indexed ImmunoSequencing transforms the Western blot from an artisanal, low-plex picture into a standardized, high-plex, quantitative assay that preserves size-resolved proteoforms. By coupling rapid microchip SDS separation and inline fractionation to DNA-barcoded immunoassays with UMI-based counting, DSI-Seq delivers isoform-aware pathway readouts from tiny inputs with built-in controls and shareable digital outputs. This capability enables rigorous kinetic mapping of signaling, proteoform-level pharmacodynamic biomarkers, and reproducible QC for cell therapy and drug programs, filling a critical gap between legacy Westerns and high-bar mass spectrometry and advancing proteoform-centric biology across immunology and oncology.

Layperson summary: Think of DSI-Seq as a modern upgrade to the classic Western blot, the lab test that shows whether a protein is present and how big it is. Instead of a fuzzy picture of bands, it separates proteins by size on a tiny chip, adds simple barcodes, and then uses DNA reading tools to count each protein like a supermarket scanner. The result is a clear table of numbers that tells you not only how much of a protein is there, but also whether it is the full version, a cut piece, or a switched-on form. It can check many proteins at once from a very small sample and gives answers in hours instead of days. This helps scientists see how cells send signals, confirm how drugs and cell therapies work, and compare results across labs with much better consistency. In short, it turns a slow, manual art into a fast, reliable measurement that can speed up discoveries and improve testing in medicine. Method: Digital Size‑Indexed ImmunoSequencing (DSI‑Seq)

Goal: preserve Western’s size information and antibody specificity, but make it quantitative, multiplexed, fast, and low input.

Core idea
Rapid microchip SDS separation of denatured proteins.
Inline fractionation of the eluting stream into many narrow size bins.

Multiplex DNA‑barcoded immunoassay in each bin.
Sequencing or digital PCR readout to count molecules per target per size bin.

Computational mapping from bin index to molecular weight using co‑run standards.

This yields a 2D matrix: targets on one axis, size bins on the other. You see isoforms and shifts exactly like a blot, but with digital counts instead of gray bands.

Why it beats Westerns

Isoform-resolved, like a blot: size bins let you separate full-length vs cleaved or shifted proteoforms.

High multiplex: 50 to 200 targets per run is realistic with barcoded antibodies.

Quantitative: counts are digital, with internal spike‑ins and UMIs for normalization and linearity.

Low sample: nanoliter fractions and single‑tube chemistry cut input requirements.

Throughput: dozens of samples per day on a benchtop box, no membranes, no overnight incubations.

Reproducibility: fixed microfluidic geometry, DNA barcodes with error correction, built‑in standards.
Workflow overview

Sample prep

Lyse in SDS buffer with protease and phosphatase inhibitors as needed.

Add a small panel of recombinant protein ladder standards spanning 10 to 250 kDa.

Microchip SDS separation

Use a disposable polymer‑sieving microfluidic chip.
Inject sample plus ladder. Separate over a short channel. Total runtime on the order of minutes.

Inline fractionation

At the channel outlet, segment the eluent into nanoliter droplets at a fixed frequency.

Each droplet receives a time‑stamp DNA barcode from a clocked side stream. The time stamp uniquely encodes the size bin because migration time maps to molecular weight. In plate‑based builds, collect sequential fractions across 64 to 128 wells instead of droplets and dispense a unique DNA bin barcode into each well.

Multiplex immunoassay with DNA barcodes

Add a cocktail of antibodies, each conjugated to a unique DNA tag that carries an antibody ID barcode, a unique molecular identifier (UMI), universal priming sites.

Two options for specificity:
Single‑epitope capture: antibodies on magnetic beads capture target in each bin. A reporter oligo is released only upon capture using a proximity‑restricted cleavage or displacement.

Proximity extension assay (PEA): two oligo‑tagged antibodies per target give a ligation‑competent DNA only when both bind the same protein. This strongly reduces off‑target.

The bin time‑stamp barcode is ligated to each target amplicon so every read is labeled with both protein identity and size bin.

Readout

PCR amplify and pool. Quantify by NGS for high plex or by multiplex digital PCR for 5 to 30 targets.

Demultiplex reads to counts per target per bin. Collapse UMIs to remove amplification bias.

Analysis

Fit ladder‑derived calibration to convert bin index to apparent kDa.

Call peaks per target to quantify proteoforms or shifts.

Normalize using: external spike‑in protein standards, per‑bin total protein dye signal, or stable reference proteins.

You can plot fraction profiles the way you read bands now, but with real numbers and confidence intervals.

Key design details

Microfluidics

Short, linear polyacrylamide sieving matrix in a molded chip.

64 to 128 fraction bins across the separation window. Effective resolution comparable to 6 to 12 percent gels.

Droplet mode or 96‑well fractionation module, depending on comfort level.

Barcodes and error control

12 to 16 nt antibody ID barcodes with Hamming distance of 3 or more.

8 to 12 nt time‑stamp barcodes for bins.

8 to 12 nt UMIs to deduplicate PCR bias.

Use a universal primer pair for all targets to simplify amplification.

Antibody conjugation

Protein A/G mediated orientation, then NHS‑PEG‑azide to click a maleimide‑bearing oligo, or use site‑specific conjugation on engineered Fc.

Validate each antibody’s linear epitope binding under 0.05 to 0.1 percent SDS or after SDS quench with cyclodextrin.

Controls

No‑antibody and isotype controls in a few bins to estimate background.

Known‑ratio mixtures of recombinant targets to test linearity.

Phosphatase‑treated lysate as negative control for phospho‑specific panels.

Performance targets to aim for

Multiplex: 100 proteins per reaction with NGS; 10 to 30 with dPCR.

Dynamic range: 4 to 5 orders of magnitude with UMIs and spike‑ins.

Sensitivity: low femtomoles per target per lane is realistic; sub‑microgram total protein per sample.

Size resolution: 5 to 10 percent across 15 to 250 kDa with 64 to 128 bins.

Hands‑on time: minimal, no membranes or overnight incubations.

These are engineering targets, not guarantees. They are chosen based on what each subsystem can credibly deliver.

Validation plan that will convince a hard skeptic

Build the separation plus fractionation stub
Run a prestained ladder through the chip.

Fractionate into 64 bins.
Confirm bin‑to‑kDa mapping with a logistic fit.

Single‑target end‑to‑end test
One antibody with an oligo tag.

Spike recombinant protein at a dilution series into a HeLa lysate.

Show linear counts versus input and a single bin peak at the expected kDa.

Isoform resolution

PARP1 cleavage in apoptosis: treat cells with staurosporine.

Detect 116 kDa full length and 89 kDa cleaved in separate bins.

PTM specificity
Phospho‑ERK1/2: stimulate and phosphatase‑treat controls.

Phospho‑specific antibody should yield bin‑aligned signal that collapses with phosphatase.

Test Multiplex panel

24‑plex T cell signaling panel: CD3ζ pY, ZAP70 pY, LAT pY, SLP76, PLCG1 pY, ERK1/2 pT/pY, AKT pS473, mTOR pS2448, 4EBP1 pT37/46, NF‑κB p65 pS536, beta‑actin as reference.

Compare fold changes to traditional Westerns and phospho‑flow as orthogonal methods.

Reproducibility

10 technical replicates. Compute CV per target per bin.
Batch‑to‑batch chip variation and antibody lot testing.
Practical build choices

Version A: plate‑basedUse capillary electrophoresis into a 96‑well plate, dispense pre‑aliquoted bin barcodes by row, perform the immuno‑DNA chemistry in wells, and read by amplicon sequencing. This is the fastest route with standard gear.

Version B: droplet‑basedMicrofluidic droplet generator assigns the bin time‑stamp barcode in flow, merges with antibody beads, then breaks emulsion for pooled amplification. Higher automation and less loss, but more microfluidic engineering.

Version C: no‑sequencer optionReplace NGS with a panel of TaqMan probes on a digital PCR platform. Lower plex but no dependence on sequencing.

Risks and mitigations

Antibody performance in residual SDSMitigation: dilute SDS below 0.05 percent, use SDS scavengers, favor linear epitope antibodies, or renature briefly before binding.

Carryover between binsMitigation: increase droplet segmentation rate or fraction dead volumes, add a short waste gap between bins, validate cross‑talk with ladder only.

Barcode cross‑talk and index hoppingMitigation: error‑correcting barcodes, unique dual‑indexing, UMI deduplication.

Off‑target binding in complex lysatesMitigation: use PEA two‑antibody logic for problematic targets, include matched isotype and competition controls.

Mapping accuracy from time to kDaMitigation: co‑run ladder and fit per run. Report apparent kDa with confidence intervals.

How it compares to current alternatives

Traditional Western: size info yes, multiplex low, quantitation poor, time long. DSI‑Seq: size info yes, multiplex high, digital counts, faster.

Capillary Western systems: automated and quantitative but limited multiplex. DSI‑Seq: retains automation but scales multiplex via barcodes.

DigiWest‑like bead fractionation: conceptually close but uses protein‑on‑bead plus fluorescent detection. DSI‑Seq: swaps fluorescence for DNA counting with UMIs and a simpler microchip separation.

Targeted mass spectrometry: high specificity and absolute quantitation, but complex setup and limited size visualization. DSI‑Seq: more accessible and preserves a blot‑like picture of proteoforms.

Kit and instrument concept

Disposable chip with sieving matrix and outlet fractionator.

Antibody panel with validated DNA barcodes and UMIs.

Bin barcode plate or droplet side‑stream mix pre‑made.

Calibration ladder proteins plus spike‑in counting controls.

Benchtop controller for voltage, flows, and droplet timing.

Software that auto‑maps bins to kDa, calls peaks, and outputs publication‑ready plots and CSVs.

When denaturation or SDS is unavoidable
If you must process denatured lysate, dilute or scavenge SDS before proximity chemistry. Many Western‑validated antibodies bind linear epitopes after SDS quench.

If two‑epitope binding fails under those conditions, switch the target to Option B capture‑and‑release for that protein while keeping proximity extension for the rest.

Failure modes to plan for, with fixes

High background without protein: incomplete nuclease inactivation or oligo self‑ligation. Fix by tightening the nuclease step and redesigning oligos with higher Hamming distance and blocked ends.

Hook effect at very high abundance: split samples or dilute to keep partition occupancy between 0.1 and 0.7.
Poor agreement between targets: recalibrate conversion efficiencies with purified standards, and check antibody pair compatibility.

Cross‑reactivity: migrate problematic targets to the capture‑and‑release format or require a competition control for acceptance.

Bottom line and cross-check with the landscape:

DSI‑Seq is a novel and feasible direction for a true “Western‑replacement” that preserves size information while giving digital, multiplexed readouts. The closest prior arts solve only parts of this: DigiWest multiplexes after SDS‑PAGE but uses bead fluorescence, not sequencing; Simple Western automates size‑resolved immunoassays but does not scale to high plex; Olink PEA and other DNA‑barcoded immunoassays are highly multiplexed but lack size resolution. Combination of microchip SDS separation + inline fractionation + bin barcoding + DNA‑counting does not appear in the literature as a unified method.

How it compares to what exists
DigiWest: SDS‑PAGE lane is sliced and eluted to barcoded beads, then probed and read by Luminex fluorescence. It retains size info and multiplexes, but detection is not digital sequencing. Your approach swaps in sequencing with UMIs and a microchip separation front‑end.

Simple Western (capillary Western): automated, quantitative, but limited multiplex and no sequencing readout.

DNA‑barcoded immunoassays: Olink PEA and ID‑seq give high plex digital counts by NGS, but they operate in solution without size resolution. You add back the Western’s unique size axis.

Microchip SDS protein sizing: commercially standard with typical ∼10 percent sizing resolution in minutes, which matches your binning targets.

Conclusion on novelty: the integration and the bin time‑stamping idea look patentable and practically differentiating.

I will begin unpacking our recent study on #MECFS using AI approaches. We think this will have a profound impact on both understanding the disease & revealing actionable targets for potential treatments. This approach can also be applied as precision medicine for chronic diseases Myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) is a complex, long-term illness that affects multiple body systems, characterized by severe, persistent fatigue that is not alleviated by rest and worsens with physical or mental exertion. It can affect anyone, regardless of age, gender, or ethnicity, though it is more commonly diagnosed in women & people aged 40-60 years.

İsrail’in Nisan’da koronayı yenmiş ve toplumsal bağışıklığa ulaştığını düşünmüştük. 2-3 ay boyunca da vaka sayıları 2 haneli rakamlara düştü haftalarca ölüm olmadı. Fakat Temmuz sonunda İsrail’de çok hızlı 4.Covid dalgası başladı, peki ne oldu? Aşılar etkisiz mi ? Bu flood da: Gerçekten günlük vaka sayıları İsrail’de 4-5 binlere çıktı, şu anda hastanede yatan covid hastaları 500-600 civarı, ölümlerde sıfırdan günlük ortalama 10-20 civarına çıktı, ve bunların yaklaşık yarısı aşılı (çift doz mRNA). Öncelikle, bu yeni dalganın ana sebebi Delta varyantı…/

Aşı karşıtları & militan takipçilerini anında engelliyorum, lakin bitmek bilmiyorlar, hakaret, saldırı & tehditleriyle…sanki organize çeteler gibi hareket ediyorlar, hayatları pahasına koronayla mücadele eden meslektaşlarımızı tedirgin ediyorlar. Bu kabul edilir bir durum değil.

https://twitter.com/zynpokmen/status/1426568964029698048

Maddi çıkar var tabii, Amerika’da bunların başı Joseph Mercola, geçen yıl sözde takviye “alternatif tıp” ürünlerinden 100 milyon dolar kazanmış! Bu şekilde popüler olan bazı doktorlar da bu işleri yapıyor, tereddütte olan insanları istismar ederek kendi etrafında topluyor.

https://twitter.com/Cetinn_Ceviz/status/1426807348929105920

Bugüne kadar koronaya karşı çoğunlukla antikorlardan (akıllı füzelerden) bahsettik ama bağışıklık sisteminin bir de komutanları ve özel kuvvetleri/keskin nişancıları (T hücreleri) var- aslında bunlar benim de 30 yıldır araştırmalarımızın çoğunu adadığım ve en sevdiğim hücreler ++ Bu T hücre komutanları virüsün parçalarını tanımazsa, iyi antikor da oluşmaz. Önemli olarak, akıllı füzelerden kaçıp hücrelere girmeyi başaran virüsleri de T hücre özel kuvvetleri arayıp bulur & yok eder. Güzel tarafı virüs mutasyonla bu T hücrelerinden öyle kolay kaçamaz😊++