Saturday, April 4, 2026

One Song, Many Tongues: On Death, Love, and What Persists


Sometimes I am amazed at how human experience finds a way of manifesting itself in poetry across different languages and cultures. While listening to this Celtic-style song "The Raven", I couldn't help but be reminded of these lines (attributed to Kabir?, my attempt at an english translation and commentary below).


کاگا سب تن کھائیو، چن چن کھائیو ماس

دو نیناں مت کھائیو، انہیں پیا ملن کی آس


آج کہ کال کہ پچھے دن، جنگل ہوئے گا باس

اوپر اوپر پھریں گے، ڈھور چرتے گھاس


پنجر پریم پرکاسیا، انتر بھیا جاس

مکھ کستوری مہکہی، بانی پھوٹی باس


کبیر کہا گر بیو، اس جوبن کی آس

ٹیسو پھول دوس چار، کھنکھر بھئے پلاس


ہاڈ جلے جیوں لاکری، کیس جلیں جیوں گھاس

سب تن جلتا دیکھ کر، بھیا کبیر داس


---


اردو تشریح:


کاگا سب تن کھائیو، چن چن کھائیو ماس: اے کوے (موت/جسم کے ٹکڑے کرنے والے)، میرے پورے جسم کو کھا لے اور چن چن کر گوشت ختم کر دے۔


دو نیناں مت کھائیو، انہیں پیا ملن کی آس: لیکن میری دونوں آنکھیں مت کھانا، کیونکہ ان میں ابھی محبوب سے ملنے کی امید باقی ہے۔


پنجر پریم پرکاسیا، انتر بھیا جاس: جب جسم کے پنجرے میں عشق روشن ہوا، تو اندر کی آگاہی بھی روشن ہو گئی۔


مکھ کستوری مہکہی، بانی پھوٹی باس: اس عشق کی وجہ سے وجود میں خوشبو پیدا ہوئی اور زبان سے مہکتی ہوئی باتیں نکلنے لگیں۔


ہاڈ جلے جیوں لاکری، کیس جلیں جیوں گھاس: آخرکار ہڈیاں لکڑی کی طرح اور بال گھاس کی طرح جل گئے۔


سب تن جلتا دیکھ کر، بھیا کبیر داس: جب پورا جسم فنا ہو گیا تو انسان حقیقی معنوں میں بندگی کی حالت تک پہنچا۔


English interpretation:


O Raven, consume this entire body, pick away every piece of flesh.

But do not touch these two eyes, for they still carry the hope of meeting the beloved.


Today or tomorrow, after a few days, this body will become a dwelling of the wilderness.

Above it, creatures will wander, and cattle will graze upon what remains.


When love illuminated this cage of the body, the inner being became awakened.

The face began to carry a musk-like fragrance, and speech itself became imbued with sweetness.


Kabir says, why cling to the illusion of youth?

Like the palash flower that blooms briefly, all beauty soon turns to dust.


Bones burn like wood, and hair like dry grass.

Seeing the whole body consumed, Kabir became truly a servant, dissolved beyond the self.


---


Lyrics --The Raven


Through the mist I see her wings,

A shadow black where twilight clings.

The raven calls the night to rise,

A song of stars that pierce the skies.


Chorus:

Oh raven, raven, take my soul,

Fly where the endless rivers roll.

Through ancient trees and whispered streams,

Oh raven, raven, carry my dreams.


Her feathers catch the moon's pale fire,

A silent witness to desire.

She knows the secrets I have lost,

Each truth I've hidden, each line I've crossed.


Pre-Chorus:

Oh, tell me where the winds have gone,

And why the earth still hums her song.


(Chorus)


Beneath her wings, the world unwinds,

She sees the threads we leave behind.

No words, no vows, no promises stay,

The raven comes and sweeps them away.


(Chorus)


==


Different languages, different traditions, but the same idea returns:


The body fades, but longing remains.

And somehow, we imagine it carried onward -- by a bird, a shadow, a song.




 

Saturday, December 13, 2025

The allegory of strange loops: A perspective on AI guided cancer research

 The allegory of strange loops: A perspective on AI guided cancer research

By Fayyaz Minhas (With AI editorial support)


I often begin my talks about my research in understanding cancer pathobiology with Plato's allegory of the cave. Prisoners sit chained to a wall, watching shadows cast by unseen objects behind them. They debate the nature of reality by analysing those shadows, unaware of the processes that generate them. In many ways, we as scientists find ourselves in a similar position. We observe measurements, images, molecular readouts, and clinical outcomes. These are shadows cast by deeper biological processes. Our task is not merely to describe the shadows, but to infer the mechanisms that generate them, so that we can intervene meaningfully in the real world.

Modern cancer research is often framed as a choice between levels of description. At one extreme lie molecular profiles such as mutations, gene expression, and epigenetic marks, treated as privileged causes of disease. At the other lie tissue images such as whole slide pathology and spatial maps, treated as rich, high dimensional summaries of downstream phenotypic effects. Increasingly, the proposed solution is multimodality: integrate everything using artificial intelligence, and the truth will emerge.

But multimodality is a tool, not a philosophy. One can be multimodal and still profoundly reductionist.

The deeper problem is not missing data, but how we, particularly as computer scientists, think about causation, explanation, and understanding in biology. Cancer and many other complex pathologies are not poorly understood because we lack modalities. They are poorly understood because we continue to force self organising, non linear systems into explanatory frames designed for linear causation.

Genes explain evolution, not construction

The phrase “it's all in the genes” is deeply ingrained in biological thinking. Richard Dawkins' The Selfish Gene is often taken as support for this idea, but this is a misunderstanding. Genetic determinism is the belief that knowing the genes is sufficient to predict biological outcomes, that DNA functions as a blueprint whose execution yields inevitable form and behaviour. Dawkins does not make this claim. His argument is precise and limited. Genes are replicators, units of evolutionary persistence. They explain why certain traits spread through populations over evolutionary time, much like memes spread culturally. They do not explain how organisms are constructed, how tissues organise, or how disease unfolds dynamically within an individual.

Philip Ball's How Life Works makes this distinction explicit. Genes are not programs or blueprints. They do not encode form directly. Biological order arises from interactions among genes, physical constraints, cellular dynamics, and environmental context. Development, robustness, and pathology are processes rather than executions of code. Form is constructed through interaction, not read out from DNA.

Yet in cancer biology, gene centred thinking quietly hardened into genetic determinism. One gene, one solution narratives flourished, sometimes implicitly and sometimes explicitly. These narratives achieved spectacular success in particular contexts, such as subsets of acute myeloid leukaemia, where single mutations can explain disease mechanisms and enable targeted therapy.

These successes were real, but they were also misleading. They encouraged the belief that cancer in general is a molecular wiring problem waiting to be debugged, perhaps one day to be edited away. Most diseases are not like this. They are not deterministic, linear, or reducible to isolated molecular causes.

Life, order, and the strange loop

My thinking here owes much to Gödel, Escher, Bach and I Am a Strange Loop by Douglas Hofstadter. I did not fully understand these books when I first read them, and I may never fully understand them. This is not a weakness of the books. As Nassim Nicholas Taleb has observed, if a book can be fully understood by a summary, it likely did not need to be written. Books that continue to yield new interpretations upon rereading are often those grappling with genuinely deep ideas.

Hofstadter introduced a way of thinking that has proven indispensable. Complex systems generate meaning and behaviour through circular causation. In a strange loop, higher level patterns emerge from lower level interactions and then act back upon them. The system becomes self referential. Escher's drawings, where hands draw one another or staircases loop back onto themselves, are visual demonstrations of this principle.

This is Hofstadter's account of consciousness, but it is also a powerful account of biology. Cancer, viewed through this lens, is not a simple accumulation of mutations. It is an emergent reorganisation of a tissue level system. It adapts, persists, and exhibits apparent goal directedness. When we say that a tumour invades or that cancer metastasises, we are not invoking intention, but we are not speaking nonsense either. We are describing emergent teleology, behaviour that appears purposeful because it arises from feedback driven organisation.

Order against entropy and the origin of emergence

Where does this apparent purpose come from in life and disease. William James Sidis, in The Animate and the Inanimate, approached this question from a thermodynamic perspective. Sidis suggested that life locally reverses the second law of thermodynamics. This is not correct as literal physics. Living systems do not violate thermodynamic laws. However, Sidis captured something enduring and important. Living systems are non equilibrium systems that maintain and propagate structure against dissipation through continuous flows of energy and information.

This sustained organisation creates feedback loops. These loops stabilise patterns, amplify certain behaviours, and suppress others. Over time, such loops give rise to structures that persist, adapt, and interact. Even pathological life does this. Cancer is not random noise. It is structured, patterned, and persistent, almost like one strange loop interacting destructively with another.

Teleology in biology does not arise from intention but from stability. Patterns that persist appear purposeful because they have survived the dynamics that would otherwise dissolve them.

Cancer emerges at the tissue level

Complex systems often exhibit properties that are more than the sum of their parts. Simple local rules can give rise to complex global behaviour, a principle explored extensively by Stephen Wolfram in A New Kind of Science. Cellular automata governed by trivial rules can suddenly cross thresholds into rich, unpredictable dynamics.

Biology appears to operate in exactly this way. Below certain organisational thresholds, behaviour remains simple and local. Above them, new levels of causation emerge.

This view is formalised in the emergent theory of cancer articulated by Sigston and Williams, who reject the sufficiency of the somatic mutation theory. They argue that cancer is an emergent system arising at the level of the functional tissue unit, not at the level of genes or individual cells.

In this framework, tissue architecture is not a downstream consequence of molecular change but a primary explanatory level. Causation is multi scale and bidirectional. Genes influence tissue organisation, but tissue organisation constrains gene expression, cellular behaviour, and evolutionary trajectories.

This relationship can be formalised statistically and information theoretically. Imaging patterns reduce uncertainty about mutations or cellular states and vice versa, revealing measurable mutual information between morphology and molecular profiles. This statistical dependence enables prediction. Certain architectural patterns correlate with aggressive disease, particular mutations, or coordinated expression programs.

But statistical dependence is not causation, nor does it explain the latent mechanisms that generate both. Diagnosing cancer from genes alone is conceptually incomplete because it ignores the constraints imposed by tissue organisation. Explaining cancer purely from morphology risks mistaking stabilised patterns for primary causes. Both errors arise from collapsing a circular system into a one way explanation.

Images are not explanations either

When I moved from bioinformatics into computational pathology, I initially thought of whole slide images as another downstream layer of molecular events. Over time, this assumption eroded. Computational pathology is too often reduced to machine learning on slides rather than understood as a way to study disease processes themselves using computation, particularly the latent causes that give rise to observed patterns.

There has been substantial work on predicting gene expression or mutations from morphology. These efforts are useful, but the predictions are never perfect. The standard explanations invoke noise, limited datasets, or weak labels. These explanations are convenient but insufficient.

There is growing evidence that mechanical forces and geometry influence gene regulation. Studies have shown that nuclear deformation can alter chromatin organisation and transcriptional programs in colorectal cancer and other systems. Shape affects transcription. If mechanical perturbation of nuclei can regulate genes, then tissue architecture cannot be treated as a passive readout. Form acts back on function, reminiscent of Escher's hands drawing one another.

This is the strange loop again. Genes shape tissue. Tissue reshapes gene expression. The system stabilises around emergent phenotypes that we observe as tumour evolution. The difficulty of predicting molecular states from images is not merely noise or dataset limitation. It reflects the fact that causal prediction is fundamentally harder than anticausal prediction, especially in systems with circular causation, as emphasised by Bernhard Schölkopf and others.

Multimodality is not the point

At this stage, it is tempting to declare multimodal integration as the solution. A more holistic picture is necessary, but multimodality can easily become a fetish. One can integrate images, transcriptomics, and proteomics while still treating one level as ground truth and the others as noisy reflections. This reproduces reductionism under the guise of completeness.

The goal is not to stack modalities. The goal is to change the explanatory frame.

Causality, not comfort

This is where Judea Pearl's The Book of Why becomes essential. Pearl's central argument is that prediction is not explanation. Correlation cannot answer what if questions. Understanding requires causal models that encode interventions, counterfactuals, and feedback.

AI in biology is often asked to be explanatory in a psychological sense, to reassure us, to pacify us, to tell stories that feel intuitive or controllable. But explanatory AI can become an anaesthetic. It can obscure rather than illuminate when it reinforces simplistic narratives.

The task is not to explain cancer in comforting terms, but to reason about it correctly, because lives depend on it.

Rethinking experimentation and evidence

This shift has implications for how we design experiments and therapies. Much of modern biology still revolves around isolating individual cells, increasingly through organoids, perturbing them with drugs, and extrapolating upward. This approach has value, but it remains limited.

As Professor David Epstein once observed, death is a non subjective assessment. Outcomes matter. Retrospective patient data, longitudinal disease pathways, tissue organisation, treatment histories, and clinical endpoints encode causal information that no isolated assay can capture alone.

Understanding cancer and designing therapies should not proceed from isolated mechanisms in isolation. It should integrate retrospective patient trajectories, tissue level organisation, clinical trial outcomes, and molecular interventions, not as parallel silos but as components of a single causal system.

Toward a new view

Images alone are not enough. Mutations alone are not enough. Multimodality alone is not enough.

Cancer is neither a genetic bug nor a visual pattern. It is a failure mode of a complex, self organising system. The actionable path forward is clear. We must adopt causal, multi scale models as the foundation of AI in cancer biology. We must treat genes, cells, tissues, patients, and treatment pathways as interacting levels rather than hierarchical outputs. We must use AI not merely to predict or explain, but to test interventions and counterfactuals across scales. We must anchor discovery in real patient trajectories rather than isolated experimental abstractions.

AI should not be a pattern recogniser or a storyteller. It should be a tool for navigating strange loops.


Sunday, June 29, 2025

What Iqbal thinks of Artificial Intelligence?

 After giving it all of Iqbal's Urdu Poetry with English Translation, I asked ChatGPT to imagine itself as Iqbal and answer interview questions. Here is the response to my question about "What would Iqbal think of Artificial Intelligence?"


Artificial Intelligence? A marvel of human thought, no doubt—a flame sparked by the same fire that once taught Adam the names of all things.

But let me ask you:

Has man mastered himself before he dares to master machines?

Intellect (‘aql) is a lamp, but without ishq—love, vision, and spiritual yearning—it is blind.

If AI is born from intellect but lacks soul, then it is but a mirror reflecting our confusion, not our clarity.

I have seen the West bend mountains with science,

But has this knowledge birthed free men?

Artificial Intelligence may extend man’s reach—but will it deepen his insight?

It may predict his wants—but can it awaken his khudi?

Use it, yes. Shape it, yes.

But do not worship it, lest you forge new idols—cold, efficient, godless.

The greatest danger is not that machines will think like humans,

but that humans will forget how to think like men with soul.

Sunday, April 28, 2024

Can ChatGPT deduce the presence of anti-CRISPR proteins?

 

Can ChatGPT deduce the presence of anti-CRISPR proteins?



I asked ChatGPT to infer the presence of anti-CRISPR proteins by relying on its ability to reason. I gave it two separate prompts. In Prompt-1 I asked it to check if the known deduction of the presence of anti-CRISPR proteins is correct or not. In Prompt-2, I just asked it to infer a logically sound deduction from given information. 

I will let you judge the answers.


Prompt-1:

You are a master logician. I will give you some examples of logically sound and unsound deductive reasonings. Then I will ask you a question for you to check if that deduction is logical or not and also help validate the deduction as well as identify and report relevant sources on the internet to generate your answer:


Logically Sound Deductive Reasoning Examples:


All dogs have ears; golden retrievers are dogs, therefore they have ears.

All racing cars must go over 80MPH; the Dodge Charger is a racing car, therefore it can go over 80MPH.

Christmas is always Dec. 25th; today is Dec. 25th, therefore it’s Christmas.



Logically Unsound Deductive Reasoning Examples:


All zebras have stripes; tigers have stripes, therefore tigers are zebras.

Fourth of July always has fireworks; today there were fireworks, therefore it must be the Fourth of July.

Carrots are orange; oranges are orange, therefore oranges are carrots.


Question:


Bacteria use CRISPR to destroy viral DNA to prevent infections

Their CRISPR array is a record of DNA of previous infections

If a bacterial contains a CRISPR array that can target some part of the bacteria’s own DNA then that bacteria then such a bacteria should self-destruct as it will destroy its own DNA

However, such bacteria do exist 


Deduction: these bacteria must contain some mechanism of inhibiting CRISPR


Response to Prompt-1 from ChatGPT:

The deduction presented here relies on understanding the mechanism of CRISPR and its function in bacteria. Let’s analyze this step-by-step.


# Deductive Reasoning Analysis

1. **Premise 1**: Bacteria use CRISPR to destroy viral DNA to prevent infections.

   - This is scientifically accurate. CRISPR-Cas systems in bacteria serve as an adaptive immune system, where they use sequences saved in their CRISPR arrays to recognize and cut specific DNA sequences, typically from viruses.


2. **Premise 2**: Their CRISPR array is a record of DNA of previous infections.

   - True. The CRISPR array consists of snippets of DNA from viruses that have previously infected the bacterium. These snippets are used to recognize and defend against future attacks by similar viruses.


3. **Premise 3**: If a bacterial cell contains a CRISPR array that can target some part of the bacteria’s own DNA, then that bacteria should self-destruct as it will destroy its own DNA.

   - Logically, if CRISPR were to target the bacterium’s own DNA, it could lead to lethal consequences for the bacterium by damaging essential genes.


4. **Observation**: However, such bacteria do exist.

   - This indicates that there are exceptions where bacteria do not self-destruct despite having CRISPR systems that could potentially target their own DNA.


5. **Deduction**: These bacteria must contain some mechanism of inhibiting CRISPR.

   - This deduction logically follows from the observation and premises. Since the bacteria that could potentially target their own DNA do not always self-destruct, it is reasonable to infer that they have mechanisms to inhibit CRISPR activity under certain conditions to prevent self-destruction.


# Validation and Sources


Research in the field of CRISPR and microbial genetics has shown that bacteria have indeed evolved mechanisms to regulate CRISPR activity to prevent self-targeting. These mechanisms can include various forms of regulation at the genetic and molecular levels to ensure that CRISPR systems do not accidentally target the bacterium’s own DNA.


For example, some bacteria have anti-CRISPR proteins that inhibit CRISPR-Cas systems, and there are specific regulatory processes that help avoid targeting the bacterium’s own DNA unless certain conditions are met. These adaptations are essential for the survival of bacteria with CRISPR systems.


Prompt-2

You are a master logician. I will give you some examples of logically sound and unsound deductive reasonings. Then I will ask you a question for you to generate a logical deduction for it and validate your response with relevant sources from the internet:


Logically Sound Deductive Reasoning Examples:


All dogs have ears; golden retrievers are dogs, therefore they have ears.

All racing cars must go over 80MPH; the Dodge Charger is a racing car, therefore it can go over 80MPH.

Christmas is always Dec. 25th; today is Dec. 25th, therefore it’s Christmas.



Logically Unsound Deductive Reasoning Examples:


All zebras have stripes; tigers have stripes, therefore tigers are zebras.

Fourth of July always has fireworks; today there were fireworks, therefore it must be the Fourth of July.

Carrots are orange; oranges are orange, therefore oranges are carrots.


Question:

Bacteria use CRISPR to destroy viral DNA to prevent infections

Their CRISPR array is a record of DNA of previous infections

If a bacterial contains a CRISPR array that can target some part of the bacteria’s own DNA then that bacteria then such a bacteria should self-destruct as it will destroy its own DNA

However, such bacteria do exist 

Deduction: To be produced by you


Response to Prompt-2 from ChatGPT: 

The logical deduction from your question can be approached by carefully analyzing the statements and their implications:


Premise 1: Bacteria use CRISPR to destroy viral DNA to prevent infections.

Premise 2: Their CRISPR array is a record of DNA from previous infections.

Premise 3: If a bacterium contains a CRISPR array that can target some part of the bacterium’s own DNA, then such a bacterium should self-destruct as it will destroy its own DNA.

Observation: However, such bacteria do exist.


Deductive Reasoning:

Conclusion: Therefore, either the CRISPR system in these bacteria must have a regulatory mechanism that prevents self-destruction when targeting its own DNA, or the bacteria have adapted in such a way that CRISPR does not initiate an auto-destructive process even when its own DNA is targeted.


Explanation:

This conclusion accounts for the coexistence of bacteria with CRISPR arrays that can target their own DNA without leading to their self-destruction, suggesting a complex interplay of genetic regulation and evolutionary adaptations that prevent the scenario described in Premise 3 from occurring universally.


Validating the Conclusion:

Now, let's validate this deduction using relevant sources from the internet to ensure that it aligns with current scientific understanding. I will search for information regarding the regulation of CRISPR systems in bacteria and mechanisms they have to avoid auto-destruction when targeting their own DNA.


The logical deduction concerning the survival of bacteria that possess CRISPR arrays capable of targeting their own DNA can be strengthened by understanding the sophisticated mechanisms bacteria use to avoid self-destruction. The key to this prevention lies in the presence of protospacer adjacent motifs (PAMs). PAMs are short DNA sequences found adjacent to the target DNA sequences in viruses (protospacers) but not in the bacterial CRISPR arrays themselves. This distinction allows the CRISPR-Cas system to differentiate between foreign DNA (which contains PAMs) and the bacterium's own DNA (which lacks PAMs). Therefore, the Cas proteins only cut DNA at sites where a PAM is present, thereby safeguarding the bacterial genome from being targeted by its own CRISPR system.


Further insight reveals that the CRISPR-Cas system employs a detailed process to ensure specificity in targeting. Initially, the system transcribes the CRISPR array into RNA, which is then processed into crRNAs. These crRNAs guide Cas proteins to target and cleave DNA that matches their sequence, but only if a PAM is present, confirming it as foreign. This sophisticated safeguard is pivotal in preventing the CRISPR system from inadvertently destroying the bacterium's own DNA, which would be catastrophic for the cell.


In summary, bacteria have evolved a highly effective CRISPR-Cas immune mechanism that includes built-in checks such as PAM recognition to ensure only foreign DNA is targeted, thus allowing them to possess CRISPR arrays that could potentially match their own DNA without resulting in self-destruction​ (Innovative Genomics Institute (IGI))​​ (Frontiers)​​ (Microbiology Note – Online Biology Notes)​.




Sunday, July 16, 2023

On Love by Fayyaz Minhas (A translation)

--

It was before the nights of this world had received their darkness. Before even the stars knew how to twinkle. The newly formed moon stumbled awkwardly in its brightly lit dress. It was not even familiar with its duty of revolving around the Earth. The world had just emerged from the house of possibilities and was void of any life as we know it. Thus it was the very beginning of this system called the cosmos. Time itself had just been morphed.

There was an alchemist in the heavens at that time who could see the future of the world. He knew that God was in possession of a cure for everything but the angels were hiding the miracle from Adam. He set himself to get hold of the miracle cure and one day he did get hold of it. 

He then started collecting the ingridents required for producing it. He got brilliance from the stars, asked the moon for its grief (of separating from the Earth). Got darkness from the night. He acquired yearning from lightning, piety from a hooriah and Passion from Jesus, son of Mary. He collected Care-freeness from God Himself and Humility from angels.

He then mixed all these ingridents with the water of life And named the final product "love!"

He spread this substance over in whole of the universe And it is love that taught stars to move, flowers to blosson and life to thrive!

--

My humble translation of Iqbal's poem "Muhabbat" (Love) in Bang-e-Dara.

 http://www.youtube.com/watch?v=ATvGVLMdcC0



 جہاں کے جوہر مضمر کا گويا امتحاں تو ہے! 


The secrets of this world cannot hide from your eyes (Iqbal)

 "Don't Ask!" by Fayyaz Minhas

--

When my gaze questions her presence on my doorstep 

this late at night 

She gives me goosebumps as answers 

What's so special about her presence? 

Don't ask! 

When I dance with her, 

My quadriplegic soul

Never misses a beat or a step

What happens to my soul around her? 

Don't ask!

Sometimes I rise heartwards to meet her

Like a rose bush trying to escape it's muddy roots

Or a cloud flying high away from its watery origin 

Or the moon, running breathless without any air

Or the smoke taking to the skies, away from its fire

Who am I? Who is she? 

Earth? Water? Fire? Air? 

Don't ask!