Home page
Quantum computer is useless
Quantum computers cannot calculate drugs.
(Fig.1) Quantum computers cannot deal with infinite patterns of the molecule-protein interactions.
Physicists try to rely on the impractical quantum approximate optimization algorithm (= QAOA ) or fraudulent quantum annealing (= D-Wave ), which is a hybrid quantum-classical algorithm (= hybrid means just classical computing ) by roughly changing the molecular docking process into a optimization problem (= which is impossible ! ) where the lowest (or highest ) energy state is made to be the solution representing the most stable molecular-protein docking state.
This form for optimization problem is called Quadratic Unconstrained Binary Optimization (QUBO) where some quantum bitstring is made to mean the lowest (or highest ) energy state or the solution based on the artificially-adjusted qubit interaction parameters (= adjusted by experimental values, Not by useless quantum mechanics unable to predict anything ) in Hamiltonian H total energy or cost function ( this-p.2 ).
The point is this hybrid QAOA or quantum annealing is useless, cannot calculate anything or predict drug discovery.
A protein or a amino acid chain can have almost infinite numbers of different 3-dimensional structures depending on freely-rotatable amino acid's bonding (= torsion ) angles.
Quantum mechanics is useless, cannot solve its Schrodinger equation nor predict any multi-electron atomic energy, much less discover drugs.
So QAOA and quantum annealing have to prepare infinite numbers of different docking patterns and interaction energies obtained from experiments between molecules and target proteins to find the most stable or lowest-energy state expressed as some bitstrings, which is impossible (= infinite qubits are necessary ).
↑ Because each protein can take almost infinite numbers of different 3-D structures due to the freedom of amino-acid rotating angles, and there are almost infinite different positional relationships between the protein and molecules in slightly different positions.
↑ Each atomic position is constrained by kinds and torsion angles of other amino acids in a protein, which need many complicated constraint conditions about protein structures, which cannot be expressed by qubits.
Furthermore, quantum annealing (= fake quantum computer ) always gives wrong answers, stuck in incorrect local energy minima, so D-Wave has to create deceptive hybrid quantum computers which are just ordinary classical computers that cannot outperform classical computers ( this or this-p.1,p.3, this-p.1-abstract ).
↑ QAOA is also a deceptive hybrid method where an ordinary classical computer must find (= optimize ) solutions (= lowest-energy equilibrium state, this-p.2-3rd-paragraph ) from the input parameters with No quantum computation in optimization problems such as traveling salesman problems.
This or this-lower-Approximation Algorithms Challenges and Limitations say
"Quantum approximation algorithms have additional challenges and limitations:
The approximations are not guaranteed to be the best possible results, "
"This means that an algorithm such as QAOA may be run again, and the results might be either closer or farther away from the optimal results." ← QAOA gives wrong answers.
"QAOA and VQE both rely on classical machine learning algorithms, which introduce their own challenges and limitations" ← No quantum computation
"Current quantum computers are too error-prone to be of any usefulness today,." ← Quantum computers are impractical
This-p.1-introduction says -- Classical beats quantum
"Well-known examples
of such quantum heuristics are quantum annealing and the quantum approximate optimization algorithm (QAOA)... Moreover, our problem-tailored
classical heuristic outperforms quantum approaches in terms of solution quality for fixed runtime."
↑ To avoid infinite qubits encoding infinite different molecular docking patterns, the useless QAOA and D-Wave have to know the right answers (= right molecular docking patterns ) in advance from the experimental observation to illegitimately reduce the numbers of qubits encoding small numbers of artificially-chosen molecular docking states, which cannot predict new drugs.
Neither quantum nor classical computers can predict drug or molecular docking with target proteins due to their needing infinite numbers of experimental molecular-protein interaction patterns.
So we should use experimentally-observed molecular-protein interactions (by using useful multi-probe atomic force microscopes ) from the beginning instead of wasting time in these impractical quantum mechanics and quantum computers (= this-figure-middle~lower ).
This fake molecular docking by QAOA ↓
p.1-right-2nd-paragraph says -- Useless quantum computers
"However, approaches based on scoring functions require precise computation of the binding energy between
proteins and ligands. Even a slight error of 6 kJ/mol can
cause significant deviations in the docking results,
and achieving such accuracy with current quantum devices is difficult"
p.2-right-1st-paragraph says -- Selecting right molecular patters
"These pharmacophores
include.. the
hydrogen-bond donor or acceptor,.. Given the constraints imposed by the
current limitations of quantum devices, our study strategically narrows its focus to the most significant pharmacophores or adopts heuristic methods for selecting pharmacophore points." ← Artificially select already-known right molecular interactions with No prediction
p.2-4th-paragraph says -- Selecting right molecular patterns
"the vertex, denoted as (vl, vp), encapsulates a pair of points, where vl denotes a vertex originating from the ligand, and vp signifies a vertex from
the protein." ← Each vertex (= each qubit ) contains already-known right combinaions of the ligand and the target-protein's molecules with No prediction
"Within this framework, each edge in the BIG, expressed as (vli, vpi ) − (vlj, vpj ), delineates two vertices on the edge that are capable of co-existing in a potential docking posture" ← The edge is the constraint linking two vertices (= each vertex expressing some combination of the ligand and the protein's atoms, which can exist or not depending on the conditions of other vertices linked by the edge line )
p.3-left-2nd-paragraph says -- Experimental data needed
"Choosing a specific ϵ value involves combining empirical observations, experimental data, and theory"
p.3-Fig.1 says -- Only 7 atoms are chosen
"For analytical purposes, three pharmacophores on the ligand and four on the protein were identified,"
p.3-right-last~p.4 says -- weight = already-known potential energy
"In this graph, each vertex is associated with a positive weight wi, indicative of its pharmacophore potential"
p.6-left-1st-paragraph says -- 6 qubits = Not a quantum computer
"We begin our numerical investigation (= just classical computer's simulation ) with a 6-qubit (= one qubit takes only 0 or 1 value, so still Not a quantum computer )
quantum system"
p.6-left-4th-paragraph says -- Already-known answers
"Given the constraints of current quantum devices,.... While the actual binding
pose remains unknown in real scenarios, a similar set
of points can be procured using drug discovery knowledge. This includes heuristics selection of ligand pharmacophore points and the use of prior knowledge of binding"
p.6-right-1st-paragraph says -- Only 6 atomic points chosen
"Given our selected pharmacophores, six possible vertices emerge in the binding
interaction graph:"
p.6-right-last~p.7 says -- Experimental potential energy used
"Table. I shows the pharmacophore potential we applied in the numerical experiments... these experience parameters are derived
from the PDBbind (= experimental ) dataset"
This paper on Pfizer's useless QAOA molecular docking ↓
p.1-abstract says -- Only classical computer used
"This paper presents a Digitized Counterdiabatic QAOA (DC-QAOA) approach to molecular docking. Simulated
quantum runs were conducted on a GPU (= classical computer ) cluster."
p.2-last~p.3 says -- Answers must be known in advance
"Warm
starting involves initializing the quantum algorithm with a solution obtained from a classical algorithm or from a previous
iteration of the quantum algorithm.... In the context
of the NISQ era, where quantum resources are limited, warm-starting represents a valuable tool for maximizing the effectiveness
of QAOA"
p.3-Methods say -- Proteins cannot be predicted
"Due to the size of target proteins, the degree of freedom is usually
not manageable. In most practice of molecular docking,... Although the protein pocket remains fixed during the sampling, the large degree of freedom
still makes it challenging to cover the space or consume resources and time to find a reasonable pose."
"In order to further simplify the complexity from degree of freedom, the binding pocket and ligands can be represented as pharmacophores which define critical non-covalent interactions between ligand and protein" ← illegitimately simplify molecular interaction by already-known answers
p.4-(2)~(4) says
"a decision variable xi ∈ {0,1} indicating exclusion/inclusion in the solution clique. We may frame a maximization objective for vertices...
to recover the maximum-weighted solution clique. Next, consider two vertices vi
, vj which do not share an edge in G. Indeed,
these vertices should not be both included in the clique. We may encode this constraint as.." ← Edge is the constraint connecting two vertices that can co-exist (= each vertex contains a pair of bound atoms from a ligand and a protein )
p.7-2nd-paragraph says -- Right answers (= atoms ) must be already known
"Here are more details on the formulation of the binding interaction graph (BIG) or docking graph, which we recreate here
for completeness: we assume that pharmacophore points on both the protein – { P1,P2,...,Pn } – and ligand compound –
{ L1,L2,...,Lm } – have been selected through expert manual selection or data-driven approaches" ← artificially select right bound atomic pairs from the already-known answers with No prediction
"We construct pairs of pharmacophores based on their possibility to bind, abstracted into a vertex as vp = (Pp,Lp′), which can have.. edges between them."
p.8-Figure 1 shows this research artificially chose docking molecules (= spheres ) in advance, which means they already knew right answers in advance with No prediction by quantum mechanics nor computers.
p.10-Figure 2 used only 14 qubits (= one qubit takes only 0 or 1 value, so still Not a quantum computer ) where some 14 bitstring was made to show the optimal state or right answer.
p.13-Discussion says -- No quantum computer used
"Looking ahead, we aim to transfer these computations to real quantum processing unit", which means still No quantum computers are used in this research.
↑ So this Pfizer latest research used only a classical computing simulation of just 14 qubits (= just 14 qubits or 14 bitsring cannot express real complicated protein-drug's infinite different interaction patterns which need infinite qubits ) encoding already-known experimental molecular-protein docking atoms with No quantum computation nor quantum mechanical prediction nor drug discovery.
Feel free to link to this site.